fix(alerts): an alert's email waits for the real commit, not a savepoint (#5807)

SQLAlchemy fires after_commit and after_rollback for a savepoint too. The
host-agent ingest releases one after recording its metrics, so an alert
raised by a sample started its email before the outer transaction
committed; twice on 2026-10-12 the outbox row hit the event's foreign key
first and the email was lost. A savepoint rolling back (a docker persist
failing) cleared every pending notification, the outer one's included.

_after_commit now runs its jobs only on the outermost commit, and a
rolled-back savepoint drops only the jobs added inside it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

.forgejo/workflows/ci.yml

@@ -14,6 +14,17 @@ on:
   # pull_request intentionally absent — push on [dev, main] already fires CI for
   # every dev commit and dev→main PRs. Single-operator repo, no fork PRs.
 
+# Serialize CI runs per ref so the publish lane never shares the runner's docker
+# daemon with another run. Two publishes racing on one daemon evict each other's
+# freshly-built image mid-push, breaking the post-build tag/push (issue #1093).
+# cancel-in-progress:false is deliberate — rule 46 requires EVERY dev/main push to
+# publish its own immutable :<sha> image, so a superseded run must still run to
+# completion (never cancelled) to emit its SHA. (Forgejo honors `concurrency:` —
+# CI-runner's build workflows use it.)
+concurrency:
+  group: ci-${{ github.ref }}
+  cancel-in-progress: false
+
 jobs:
   # Fast-fail lint lane. ruff is pre-installed in the ci-python image, so this
   # runs with NO dependency install and surfaces lint bounces in seconds.
@@ -58,7 +69,12 @@ jobs:
           POSTGRES_USER: steward
           POSTGRES_PASSWORD: steward
           POSTGRES_DB: steward
+        # The data directory lives in RAM (#5746): postgres images declare
+        # VOLUME for it, so without a tmpfs each run left an anonymous volume
+        # that act_runner never removes, and 2,488 of them used up the runner
+        # host's inodes.
         options: >-
+          --tmpfs /var/lib/postgresql/data
           --health-cmd "pg_isready -U steward"
           --health-interval 10s
           --health-timeout 5s
@@ -69,14 +85,24 @@ jobs:
         # act_runner (swarm-runner) puts service containers on the default bridge
         # with no embedded DNS, so the hostname `postgres` won't resolve — we
         # discover the service container's bridge IP via the docker socket. The
-        # job name `integration` (no underscore) is the docker-ps name filter.
+        # job container and its Postgres share a task-number name prefix.
         run: |
           set -euxo pipefail
           echo "=== container landscape (diagnostic for filter scoping) ==="
           docker ps -a --format '{{.ID}} {{.Image}} -> {{.Names}}'
           echo "=== end landscape ==="
-          PG=$(docker ps --filter "name=integration" --filter "ancestor=postgres:16-alpine" -q | head -n1)
-          test -n "$PG"
+          # The runner's daemon is shared: another repo's integration job can be
+          # running beside this one, and `name=integration` matches its Postgres
+          # too (run 8358 took the wrong one and every test failed to connect).
+          # Scope to this job's own task: act_runner names every container of a
+          # task GITEA-ACTIONS-TASK-<n>-…, so read <n> from our own container
+          # (its id is in the /etc/hostname bind mount) and require one match.
+          SELF=$(grep -o '/containers/[0-9a-f]\{64\}/' /proc/self/mountinfo | head -n1 | cut -d/ -f3 || true)
+          SELF=${SELF:-$(hostname)}
+          TASK=$(docker inspect -f '{{.Name}}' "$SELF" | grep -o 'GITEA-ACTIONS-TASK-[0-9]*-')
+          test -n "$TASK"
+          PG=$(docker ps --filter "name=$TASK" --filter "ancestor=postgres:16-alpine" -q)
+          test "$(printf '%s\n' "$PG" | grep -c .)" -eq 1
           PG_IP=$(docker inspect -f '{{range .NetworkSettings.Networks}}{{.IPAddress}}{{end}}' "$PG")
           test -n "$PG_IP"
           export STEWARD_DATABASE_URL="postgresql+asyncpg://steward:steward@$PG_IP:5432/steward"
@@ -111,19 +137,154 @@ jobs:
         # 401s with it. REGISTRY_TOKEN is a repo Actions secret holding a
         # Forgejo token scoped read:package + write:package.
         run: echo "${{ secrets.REGISTRY_TOKEN }}" | docker login git.fabledsword.com -u bvandeusen --password-stdin
-      - name: Build + push (commit-SHA always; :dev on dev, :latest on main)
+      - name: Build (commit-SHA always; :dev on dev, :latest on main)
         run: |
-          set -euxo pipefail
+          set -eux
+          IMAGE=git.fabledsword.com/bvandeusen/steward
+          # The moving tag for this ref: dev→:dev, main→:latest (rule 46). main IS
+          # the production line, so :latest tracks main's tip; there is no :main.
+          MOVING=""
+          if [ "${{ github.ref }}" = "refs/heads/dev" ]; then
+            MOVING="dev"
+          elif [ "${{ github.ref }}" = "refs/heads/main" ]; then
+            MOVING="latest"
+          fi
+          # Tag BOTH targets at build time (mirrors CI-runner's build workflows):
+          # one `docker build -t :<sha> -t :<moving>` points both tags at the image
+          # atomically, so we never run a post-push `docker tag` of an image a
+          # concurrent run could have evicted from the shared daemon (issue #1093).
+          if [ -n "$MOVING" ]; then
+            docker build -t "$IMAGE:${{ github.sha }}" -t "$IMAGE:$MOVING" .
+          else
+            docker build -t "$IMAGE:${{ github.sha }}" .
+          fi
+
+      # Rule 164's check, BETWEEN build and push (rule 177): an image that
+      # can't boot, migrate and serve with no internet never reaches a tag.
+      # An --internal network lets it reach its own Postgres and nothing else;
+      # --network none can't work for an app that needs a database to start.
+      # Embedded DNS works on a user-defined network, unlike the services:
+      # containers above (rule 79), so the app finds Postgres by name.
+      - name: The image boots with no internet
+        env:
+          SMOKE: steward-smoke-${{ github.run_id }}
+        run: |
+          set -eux
+          IMAGE=git.fabledsword.com/bvandeusen/steward:${{ github.sha }}
+          docker network create --internal "$SMOKE-net"
+          docker run -d --name "$SMOKE-pg" --network "$SMOKE-net" \
+            --tmpfs /var/lib/postgresql/data \
+            -e POSTGRES_USER=steward -e POSTGRES_PASSWORD=steward -e POSTGRES_DB=steward \
+            postgres:16-alpine
+          # Waits with deadlines (rule 156); pg_isready, not /dev/tcp, which is
+          # a bash builtin and never succeeds under sh.
+          ready=""
+          for _ in $(seq 1 60); do
+            if docker exec "$SMOKE-pg" pg_isready -U steward -d steward -q; then ready=1; break; fi
+            sleep 2
+          done
+          test -n "$ready" || { echo "FATAL: postgres not ready within 120s"; exit 1; }
+
+          docker run -d --name "$SMOKE-app" --network "$SMOKE-net" \
+            -e "STEWARD_DATABASE_URL=postgresql+asyncpg://steward:steward@$SMOKE-pg:5432/steward" \
+            -e STEWARD_SECRET_KEY=smoke-test-placeholder \
+            "$IMAGE"
+          # The image has no curl; its own Python fetches. Prints the body.
+          fetch() {
+            docker exec "$SMOKE-app" python -c 'import sys, urllib.request
+          r = urllib.request.urlopen("http://127.0.0.1:5000" + sys.argv[1], timeout=5)
+          sys.stdout.write(r.read().decode())' "$1"
+          }
+          # Migrations run before the server starts, so /health answering
+          # means the schema applied against a real database.
+          up=""
+          for _ in $(seq 1 60); do
+            if fetch /health >/dev/null 2>&1; then up=1; break; fi
+            sleep 2
+          done
+          if [ -z "$up" ]; then
+            echo "FATAL: Steward never answered /health with no internet access"
+            docker logs "$SMOKE-app" 2>&1 | tail -80
+            exit 1
+          fi
+          fetch /health | grep -q '"ok"'
+
+          # tini is PID 1, so zombies are reaped and SIGTERM reaches Steward.
+          test "$(docker exec "$SMOKE-app" cat /proc/1/comm)" = "tini"
+
+          # The zone database is in the image, so a reader's zone resolves
+          # instead of quietly falling back to UTC (milestone 490).
+          docker exec "$SMOKE-app" python -c 'from zoneinfo import ZoneInfo; ZoneInfo("America/New_York")'
+
+          # A fresh database lands on first-run setup: a real page, its assets
+          # served by Steward itself, and no third-party host in it (rule 155).
+          fetch / > /tmp/smoke-page.html
+          grep -qi "<!doctype html" /tmp/smoke-page.html
+          grep -q "/static/" /tmp/smoke-page.html
+          if grep -Eqi 'https?://(fonts\.|cdn\.|unpkg\.|cdnjs\.|ajax\.)' /tmp/smoke-page.html; then
+            echo "FATAL: the page references a third-party host"
+            grep -Eoi 'https?://[^"'"'"' ]+' /tmp/smoke-page.html
+            exit 1
+          fi
+
+          # A backup made by the image restores into an empty database, and
+          # Steward starts on it (milestone 507). It carries configuration
+          # (the marker setting) and leaves metrics' rows out.
+          sql() { docker exec "$SMOKE-pg" psql -U steward -v ON_ERROR_STOP=1 -tA "$@"; }
+          sql -d steward -c "INSERT INTO app_settings (key, value_json, updated_at) VALUES ('smoke.backup', '\"kept\"', now())"
+          sql -d steward -c "INSERT INTO plugin_metrics (id, source_module, resource_name, metric_name, value, recorded_at) VALUES ('smoke-metric', 'smoke', 'smoke', 'smoke', 1, now())"
+          docker exec -u app "$SMOKE-app" steward backup --to /tmp/backups
+          sql -d steward -c "CREATE DATABASE restored"
+          docker exec -u app \
+            -e "STEWARD_DATABASE_URL=postgresql+asyncpg://steward:steward@$SMOKE-pg:5432/restored" \
+            "$SMOKE-app" sh -c 'steward restore /tmp/backups/steward-*.tar'
+          test "$(sql -d restored -c "SELECT value_json FROM app_settings WHERE key = 'smoke.backup'")" = '"kept"'
+          test "$(sql -d restored -c "SELECT count(*) FROM plugin_metrics")" = "0"
+          test "$(sql -d restored -c "SELECT count(*) FROM alembic_version")" -ge 1
+          docker run -d --name "$SMOKE-restored" --network "$SMOKE-net" \
+            -e "STEWARD_DATABASE_URL=postgresql+asyncpg://steward:steward@$SMOKE-pg:5432/restored" \
+            -e STEWARD_SECRET_KEY=smoke-test-placeholder \
+            "$IMAGE"
+          up=""
+          for _ in $(seq 1 60); do
+            if docker exec "$SMOKE-restored" python -c 'import urllib.request
+          urllib.request.urlopen("http://127.0.0.1:5000/health", timeout=5)' >/dev/null 2>&1; then up=1; break; fi
+            sleep 2
+          done
+          if [ -z "$up" ]; then
+            echo "FATAL: Steward never started on the restored database"
+            docker logs "$SMOKE-restored" 2>&1 | tail -80
+            exit 1
+          fi
+
+          # A stop is a clean shutdown: SIGTERM through tini to Hypercorn, which
+          # finishes within its graceful timeout and exits 0.
+          docker stop -t 30 "$SMOKE-app"
+          code="$(docker inspect -f '{{.State.ExitCode}}' "$SMOKE-app")"
+          test "$code" = "0" || { echo "FATAL: Steward exited $code on SIGTERM"; exit 1; }
+
+      - name: Push
+        run: |
+          set -eux
           IMAGE=git.fabledsword.com/bvandeusen/steward
-          docker build -t "$IMAGE:${{ github.sha }}" .
           docker push "$IMAGE:${{ github.sha }}"
           if [ "${{ github.ref }}" = "refs/heads/dev" ]; then
-            docker tag "$IMAGE:${{ github.sha }}" "$IMAGE:dev"
             docker push "$IMAGE:dev"
           elif [ "${{ github.ref }}" = "refs/heads/main" ]; then
-            docker tag "$IMAGE:${{ github.sha }}" "$IMAGE:latest"
             docker push "$IMAGE:latest"
           fi
+
+      - name: Smoke logs
+        if: always()
+        run: docker logs "steward-smoke-${{ github.run_id }}-app" 2>&1 | tail -60 || true
+
+      # A failed smoke must not leave a Postgres and an app on the shared daemon.
+      - name: Smoke clean-up
+        if: always()
+        run: |
+          docker rm -fv "steward-smoke-${{ github.run_id }}-app" "steward-smoke-${{ github.run_id }}-restored" \
+            "steward-smoke-${{ github.run_id }}-pg" >/dev/null 2>&1 || true
+          docker network rm "steward-smoke-${{ github.run_id }}-net" >/dev/null 2>&1 || true
       - name: Prune dangling layers
         if: always()
         run: docker image prune -f

Dockerfile

@@ -1,4 +1,5 @@
-FROM python:3.13-slim
+# The Python CI tests on (ci-python:3.14), so the image runs what was tested.
+FROM python:3.14-slim
 
 RUN DEBIAN_FRONTEND=noninteractive apt-get update \
     && DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends \
@@ -10,6 +11,17 @@ RUN DEBIAN_FRONTEND=noninteractive apt-get update \
         # sshpass — Ansible shells out to it for first-contact password SSH
         # (provisioning a fresh host before the managed key is installed)
         sshpass \
+        # git — clones and pulls Ansible git sources (Settings > Ansible)
+        git \
+        # tini — PID 1: reaps the zombies ssh and ansible leave behind and
+        # forwards SIGTERM to Steward so a stop is a graceful shutdown
+        tini \
+        # tzdata — the zone database, so times show in each reader's own zone
+        # (milestone 490); Debian's slim base no longer carries it
+        tzdata \
+        # postgresql-client-17 — pg_dump and pg_restore for Steward's own
+        # backups (milestone 507); 17 reads and writes a Postgres 16 server
+        postgresql-client-17 \
     && rm -rf /var/lib/apt/lists/*
 
 # Upgrade pip to suppress the version notice
@@ -19,10 +31,11 @@ WORKDIR /app
 
 COPY pyproject.toml .
 COPY steward/ steward/
-# Bundle the extras whose first-party plugins ship in the image: [ansible] for
-# the playbook runner, [snmp] (pysnmp) for the SNMP poller. Without them those
-# bundled plugins load but silently no-op for want of their runtime dependency.
-RUN pip install --no-cache-dir '.[ansible,snmp]'
+# Bundle the extras whose features ship in the image: [ansible] for the
+# playbook runner, [snmp] (pysnmp) for the SNMP poller, [ldap] (ldap3, pure
+# Python) for LDAP sign-in under Settings -> Auth. Without them those features
+# load but fail for want of their runtime dependency.
+RUN pip install --no-cache-dir '.[ansible,snmp,ldap]'
 
 COPY alembic.ini .
 # First-party plugins ship inside the image (bundled root at /app/plugins).
@@ -36,10 +49,10 @@ RUN chmod +x /entrypoint.sh
 # so the entrypoint can fix up /data permissions if needed, then drops to
 # 'app' (uid 1000) via gosu before launching steward.
 RUN useradd -m -u 1000 app \
-    && mkdir -p /data/playbook_cache /data/plugins \
+    && mkdir -p /data/ansible/sources /data/ansible/runs /data/ansible/playbooks /data/plugins \
     && chown -R app:app /data /app
 
 EXPOSE 5000
 
-ENTRYPOINT ["/entrypoint.sh"]
+ENTRYPOINT ["/usr/bin/tini", "--", "/entrypoint.sh"]
 CMD ["steward", "--host", "0.0.0.0", "--port", "5000"]

README.md

@@ -119,7 +119,10 @@ See `docs/plugins/` for a full plugin development guide.
 | `docs/core/monitors.md` | Ping and DNS monitors: probe logic, metrics emitted, data models |
 | `docs/core/alerting.md` | Alert rules, state machine, notification channels, template variables |
 | `docs/core/ansible.md` | Playbook sources, run lifecycle, SSE streaming |
+| `docs/core/backups.md` | Steward's daily backup to an S3 bucket (Fabled Archives), what it holds, and `steward restore` |
 | `docs/plugins/overview.md` | Plugin system: how loading works, plugin.yaml schema, required exports |
 | `docs/plugins/writing-a-plugin.md` | Step-by-step plugin development guide |
 | `docs/plugins/traefik.md` | Traefik plugin: config, metrics, alert rule examples |
+| `docs/plugins/prometheus.md` | Prometheus plugin: scraping any /metrics endpoint, how series are stored, the Fabled Archives recipe |
+| `docs/plugins/prometheus-widgets.md` | Prometheus widget templates: writing, sharing and copying one in, every block, the format reference |
 | `docs/reference/code-map.md` | Where every key function, model, and route lives in the codebase |

ci-requirements.md

@@ -46,7 +46,7 @@ git.fabledsword.com/bvandeusen/ci-python:3.14
 
 - Steward has **no frontend and no Redis/Celery** (no external workers), so there
   is no frontend-build lane and the only service container is Postgres.
-- Integration uses Forgejo Actions `services:` + a socket-discovered bridge IP
+- Integration uses Fabled-Git Actions `services:` + a socket-discovered bridge IP
   because `act_runner` (swarm-runner v0.6+) puts services on the default bridge
   with no embedded DNS. FabledCurator's `ci.yml` is the canonical example of the
   pattern; Steward's is the same shape minus Redis and minus sharding.

compose.deploy.yml

@@ -39,8 +39,7 @@ services:
       # mount, pin STEWARD_SECRET_KEY (env or a docker secret) instead, so the key
       # never depends on filesystem ownership being correct on every node.
     depends_on:
-      db:
-        condition: service_healthy
+      - db
     restart: unless-stopped
 
   db:
@@ -54,11 +53,9 @@ services:
     volumes:
       - pgdata:/var/lib/postgresql/data
     # No host port published — db is reachable only on the backend network.
-    healthcheck:
-      test: ["CMD-SHELL", "pg_isready -U steward"]
-      interval: 5s
-      timeout: 5s
-      retries: 5
+    # No healthcheck: a slow disk (NFS) failed pg_isready and the orchestrator
+    # killed a healthy database mid-write (#5275). Steward waits for the
+    # database itself, at startup and per request (steward/database.py).
     restart: unless-stopped
 
 networks:

config.example.yaml

@@ -13,6 +13,14 @@
 database:
   url: "postgresql+asyncpg://steward:password@localhost/steward"
 
+  # Optional: seconds to wait at startup for the database to become reachable
+  # before giving up (default: 60). Steward retries with backoff rather than
+  # crashing when the DB is merely slow to arrive — replaying WAL after an
+  # unclean shutdown, or container DNS not yet resolving after a host reboot.
+  # Raise it if your database is routinely slower than this to accept
+  # connections. Env var: STEWARD_DB_CONNECT_TIMEOUT
+  # connect_timeout: 60
+
 # Optional: override the auto-generated secret key.
 # If not set, a key is auto-generated on first run and saved to /data/secret.key.
 # secret_key: "change-me-to-a-random-string"

docker-compose.yml

@@ -27,8 +27,7 @@ services:
       - STEWARD_SECRET_KEY=dev-secret-key-change-me
       - PYTHONPATH=/app
     depends_on:
-      db:
-        condition: service_healthy
+      - db
     restart: unless-stopped
 
   db:
@@ -42,11 +41,9 @@ services:
       - "5432:5432"
     volumes:
       - pgdata:/var/lib/postgresql/data
-    healthcheck:
-      test: ["CMD-SHELL", "pg_isready -U steward"]
-      interval: 5s
-      timeout: 5s
-      retries: 5
+    # No healthcheck: a slow disk (NFS) failed pg_isready and the orchestrator
+    # killed a healthy database mid-write (#5275). Steward waits for the
+    # database itself, at startup and per request (steward/database.py).
     restart: unless-stopped
 
 volumes:

docs/architecture.md

@@ -44,6 +44,30 @@ All routes are Quart Blueprints registered in `app.py`:
 
 Plugin blueprints are mounted automatically by `load_plugins()` using the plugin directory name as the URL prefix.
 
+`/ai-access/` (`ai_access_bp`, `steward/ai_access/routes.py`) is where a signed-in user manages their API tokens.
+
+### `/mcp`: outside the blueprint pipeline
+
+`mount_mcp(app)` (`steward/mcp/server.py`) wraps `app.asgi_app`. A request to
+`/mcp` is handed to the MCP SDK's streamable-HTTP app before Quart sees it, so
+no blueprint, session cookie or `before_request` hook runs for it. Each
+request then goes through these steps:
+
+1. **Database gate.** If the database is unreachable, the request gets a 503
+   with a readable message.
+2. **Bearer resolution.** `api_tokens.resolve_token()` turns the token into a
+   `Caller` (user, effective role, token id). A missing, revoked or expired
+   token gets a 401.
+3. **Caller context.** The `Caller` and the app are set in ContextVars
+   (`steward/mcp/_context.py`). Tools read them through `current_caller()` and
+   `mcp_app()`, because Quart's `current_app` and `session` do not exist there.
+4. **Role gate.** `StewardMCPServer.refusal()` hides and refuses tools above
+   the caller's role, and plugin tools whose plugin is disabled.
+
+The SDK's session manager runs in Quart's `before_serving`/`after_serving`
+lifecycle. The server is stateless: no MCP session outlives a request. See
+[MCP server](core/mcp.md) for the security model and the tools.
+
 ---
 
 ## Scheduler
@@ -63,6 +87,48 @@ Core tasks registered in `app._task_registry`:
 
 ---
 
+## Metric Store
+
+Every plugin and monitor records into one store keyed by (source, resource, metric):
+- `plugin_metrics` holds raw samples for `metrics.retention.raw_days` (7 by default).
+- `plugin_metric_latest` holds each key's newest value.
+- `data_cleanup` rolls raw samples older than the raw window into `plugin_metrics_hourly`, kept for `metrics.retention.rollup_days` (90 by default).
+
+Each rolled-up hour keeps three figures:
+- `value_avg`: the hour's average.
+- `value_max`: its peak.
+- `value_p95`: the level it held for all but its briefest 5%, about 3 minutes.
+
+A two-minute spike raises the peak and leaves the p95 low; an hour spent at that level raises both. Learned normals use this to tell a recurring spike from a held level. Hours rolled up before the p95 was kept have it NULL; read those as their average.
+
+### Normals and anomalies
+
+A metric a plugin registers with `learn_normal()` gets a normal per hour of the day, then per hour of the week once three weeks of history exist (`metric_norms`, relearned hourly by `metric_norms`; `steward/core/norms.py`).
+
+Every minute, `metric_anomalies` averages each learned key's last 15 minutes and compares it with the normal for that hour. A level above the usual hold ×1.25, and above the metric's floor, opens an anomaly (`steward/core/anomalies.py`):
+- `metric_anomalies` holds one row per item to review. A repeat while the row still waits joins it as another occurrence (`metric_anomaly_spans`).
+- A row is `to_review` until someone marks it `normal` or `real` (Alerts → Anomalies, or MCP `review_anomaly`). An ended row nobody reviews within `anomalies.review_days` (14) becomes `lapsed`.
+- Learning skips every hour an occurrence covers unless its anomaly was marked `normal`, so a new level is never learned without review.
+- `data_cleanup` prunes reviewed and lapsed rows at `data.retention_days`; a row still to review is never pruned.
+
+What learns a normal, and the floor each departure must clear (Settings → Thresholds → Normals, `norms.floor.*`; a floor may be a function, `floor_setting()`, read at each pass):
+
+| Kind | Metric | Floor (default) |
+|---|---|---|
+| Host CPU | `host_agent` / `cpu_pct` | 10 points |
+| Host memory | `host_agent` / `mem_used_pct` | 5 points |
+| Host load per core | `host_agent` / `load_pct` (load_1m ÷ cores × 100, recorded at ingest) | 10 points |
+| Container CPU | `docker` / `cpu_pct` (100% is one core) | 10 points of one core |
+| Container memory | `docker` / `mem_pct` (of its limit, or of host RAM when it has none) | 5 points |
+| Monitor response time | `<type>` / `response_ms` | 100 ms |
+| Router 4xx share, p95 | `traefik` / `error_rate_4xx_pct`, `latency_p95_ms` | 2 points, 50 ms |
+
+Host metrics learn with `parts=False`: the host learns, its cores (`<host>:core<N>`) don't. Disk space, a monitor that is down, an agent gone quiet and a router's 5xx share are failures and keep their fixed cutoffs.
+
+Pages grade by the normal through `anomalies.graded()` (hosts' `host_grades` and `worst_reading(normals=)`, `docker.service.usage(thresholds=, norms=)`, `monitors.service.latency_grade`): a reading past its cutoff but within its normal reads healthy and says what is usual at this hour. The host, monitor, container and Traefik pages list their anomalies with `_anomaly_card.html`.
+
+---
+
 ## Database Session Pattern
 
 `app.db_sessionmaker` is an SQLAlchemy `async_sessionmaker`. All DB access follows this pattern:

docs/core/alerting.md

@@ -8,7 +8,7 @@ Alert rules evaluate every metric that flows through `record_metric()`. There is
 
 When any monitor or plugin calls `record_metric()`:
 
-1. The metric value is written to the `plugin_metrics` table
+1. The metric value is written to the `plugin_metrics` table, and `plugin_metric_latest` moves to it if it is the newest sample of that metric
 2. All enabled `AlertRule` rows matching `(source_module, resource_name, metric_name)` are loaded
 3. Each matching rule is evaluated against the new value
 4. If a state transition occurs, an `AlertEvent` row is written
@@ -32,18 +32,25 @@ Notifications are sent via `asyncio.create_task()` after the function returns, s
 Each `AlertRule` has one associated `AlertState` row. The state transitions are:
 
 ```
-inactive ──(breached)──► pending ──(consecutive count met)──► FIRING
-                │                                               │
-                └──(recovered)──► inactive (no notification)   │
-                                                               │
-                                          FIRING ──(recovered)──► RESOLVED ──► inactive
-                                          FIRING ──(acknowledged)──► ACKNOWLEDGED
-                                          ACKNOWLEDGED ──(recovered)──► RESOLVED ──► inactive
-                                          ACKNOWLEDGED ──(re-breached)──► FIRING
+inactive ──(breached)──► pending ──(N breaching checks in a row)──► FIRING
+    ▲                      │                                          │
+    └──(an OK check)───────┘                                          │
+                                 FIRING ──(acknowledged)──► ACKNOWLEDGED
+                                 FIRING / ACKNOWLEDGED ──(M OK checks in a row)──► RESOLVED ──► inactive
 ```
 
+N is the rule's consecutive checks, M its checks to clear. An acknowledgement
+holds for as long as the breach lasts: only a resolve clears it, and a breaching
+check never sends an acknowledged alert back to firing.
+
 `RESOLVED` is transient — the evaluator writes the event and immediately sets state back to `inactive` within the same transaction. `RESOLVED` never persists as a final state in the DB.
 
+Every transition is an `AlertEvent`, the history the Alerts → History page and
+MCP `get_alert_history` read. A transition a person made carries them in
+`actor_id`: an acknowledgement (firing → acknowledged), or a rule edit, disable
+or enable that reset the alert (→ resolved, with a note such as "rule
+disabled"). Transitions caused by readings, or by Steward itself, have none.
+
 ### State Definitions
 
 | State | Meaning |
@@ -51,7 +58,7 @@ inactive ──(breached)──► pending ──(consecutive count met)──
 | `inactive` | Threshold not breached |
 | `pending` | Threshold breached but consecutive count not yet met; no notification sent |
 | `firing` | Consecutive count met; FIRING notification sent |
-| `acknowledged` | Operator acknowledged; suppresses repeat notifications; auto-clears on recovery |
+| `acknowledged` | Someone has seen it; no further notifications; clears when the readings recover |
 | `resolved` | Transient — notification sent, immediately transitions to `inactive` |
 
 ---
@@ -68,6 +75,39 @@ Alert rules are created in the UI at `/alerts/`. Each rule requires:
 - **Threshold** — numeric value
 - **Consecutive failures required** — how many consecutive breaches before FIRING (default 1)
 
+The form offers only what has reported, read from `plugin_metric_latest`.
+For the host agent, a **Host** is picked first, then a **Part** of it: "The
+whole host", then its mounts, CPU cores, disks and so on, under a heading for
+each kind. A part that reports nothing a rule can watch, such as an interface's
+traffic, isn't listed. The **Metric** list holds only what the picked resource
+reports. Under it, a preview shows the metric's current reading and its last
+24 hours (lowest to highest, with an hourly sparkline; it loads just after the
+page, since a day of samples can take a while to read), so the threshold can be
+set against real numbers.
+
+### Rules that cover hosts
+
+A rule on `host_agent` can cover **every host**, or **every host in an
+inventory group**, instead of naming one resource (#5464). Inventory groups
+are Steward's grouping system: a host is in a group through its linked
+Ansible target.
+
+A covering rule has no state of its own. Steward keeps one ordinary rule per
+covered host under it, named `<rule>: <host>`, so each host's alert keeps its
+own state, history and acknowledgement. Each minute (the `alert_reconcile`
+task), and whenever the rule is saved:
+
+- a host that has reported in the last 7 days, and is covered, gets a rule;
+- a host that is no longer covered loses its rule, and that rule's history;
+- the hosts' rules take the covering rule's condition, checks, notification
+  setting, action and on/off.
+
+A host's rule is changed through its covering rule; editing, disabling or
+deleting it alone is refused. What a rule covers is fixed once it exists, but
+a group rule can move to another group. Over MCP, `create_alert_rule` takes
+`covers` ("all_hosts" or "group") and `group`, and `list_alert_rules` shows
+each covering rule's `hosts` and each host's rule's `part_of`.
+
 ### Available Metrics by Module
 
 | `source_module` | `metric_name` | Description |
@@ -83,6 +123,24 @@ Alert rules are created in the UI at `/alerts/`. Each rule requires:
 | `traefik` | `latency_p95_ms` | Approximate p95 latency (ms) |
 | `traefik` | `latency_p99_ms` | Approximate p99 latency (ms) |
 
+#### Steward itself
+
+Source `steward`, resource `process`, recorded once a minute by the
+self-monitor tick (`steward/core/self_monitor.py`; the ingest tally is in
+`steward/core/health.py`). Settings → System (`/settings/system/`, admin)
+shows each one now and over the last day, with a verdict above that names
+failed agent posts, a gap in watching, or open files near the limit.
+
+| `metric_name` | Description |
+|---|---|
+| `ingest_posts` | Host-agent posts Steward took in the last minute |
+| `ingest_failed` | Posts in the last minute that didn't land (handler raised, or the agent gave up) |
+| `ingest_slowest_ms` | The slowest post in the last minute; agents give up at 10 s |
+| `db_roundtrip_ms` | A `SELECT 1`, connection checkout included |
+| `unwatched_secs` | The gap before this tick when it exceeds `watch.GAP_S` (a deploy or an outage); 0 when unbroken |
+| `cleanup_secs` | How long the last hourly retention and rollup run took |
+| `open_fds` / `open_fds_pct` | Steward's open files, and as a share of its limit (absent without `/proc`) |
+
 ---
 
 ## Notifications
@@ -104,6 +162,43 @@ Configured at `/settings/` under SMTP. Settings:
 
 Email is skipped if `smtp.host` is empty.
 
+Alerts that arrive together go out as one email. The first alert opens a
+2-minute window. When it closes, each recipient gets one message listing every
+alert in the window, oldest first, in their own zone, with a subject like
+"[Steward] 3 alerts: 2 firing, 1 resolved". One alert on its own keeps its
+usual email.
+
+Every email Steward sends (alerts, the weekly report, invites, password
+resets and the test buttons) waits in the database (`email_outbox`,
+`core/outbox.py`) until the mail server accepts it, so nothing waiting is lost
+when Steward restarts: the sender picks it up at boot. The sender runs every 15
+seconds and settles each email on the server's answer:
+
+- **Accepted:** sent.
+- **Refused** (an SMTP 5xx, such as "no such mailbox"): failed at once, with
+  the server's words. Trying again would only repeat the refusal.
+- **Temporary** (no connection, a timeout, a 4xx): tried again after 30 s,
+  1, 2, 5, 10 and then every 15 minutes. An email the server still hasn't
+  taken an hour after its first try is failed with the last error.
+
+Once every email carrying an alert has settled, that alert's event records
+whether the email went, and to whom it didn't.
+
+Where to see it: each alert's row in **Alerts → History** says what its email
+is doing (gathering, with when it sends; retrying, with the attempt, the next
+try and the server's last answer; sent; failed, with who wasn't reached and
+why), and `get_alert_history` over MCP reports the same. **Settings → System →
+Outgoing email** lists every email of the last 7 days not yet sent, failures
+first, and the page's verdict names failures and any email still retrying
+after 5 minutes.
+
+An email a person is waiting on is tried at once instead of on the next tick:
+the weekly report's **Send now**, an invite, and the test buttons, so the page
+can say what the server answered. A report or invite the server couldn't take
+yet stays queued and is retried as above. A test (the Settings → Notifications
+test email, a rule's **Send test**) is recorded but never retried: its one
+answer is the result.
+
 ### Webhook
 
 Configured at `/settings/` under Webhook. The body is a Jinja2 template rendered to JSON. Content-Type is always `application/json`. If the rendered template is not valid JSON, the delivery is logged as failed and no request is sent.
@@ -124,7 +219,29 @@ Default template (Discord-compatible):
 {"content": "**{{ alert.state }}** — {{ alert.resource }} — {{ alert.rule_name }} ({{ alert.metric }} = {{ alert.value }})"}
 ```
 
-Webhook is skipped if `webhook.url` is empty.
+Webhook is skipped if `webhook.url` is empty. Webhooks are not gathered into
+bursts: each alert posts on its own, at once.
+
+### Send test
+
+**Send test** at the top of a rule's page (operator role), or `send_test_alert` over MCP,
+sends a `TEST` alert for that rule now, by email and webhook, the same way a
+real alert would go. It uses the rule's name and condition and the latest
+reading of what the rule watches. A rule that covers hosts uses its first
+host. Each channel's result shows under the button. Nothing fires, the rule's
+history doesn't change, and the test is recorded in the audit log as
+`alert_rule.tested`.
+
+### On the dashboard
+
+The **Alerts** widget (`/alerts/widget`) sits beside the Overview in the
+recommended layout. It opens with a verdict ("2 firing, 1 silent." or
+"Nothing firing: all 12 rules are quiet."). Below that it lists what is
+firing or silent: each rule, what it watches, its reading now against what
+it fires on, and since when. Acknowledged alerts follow, with who
+acknowledged them and when. A line appears when alert email has failed in
+the last 7 days or is still retrying, linking admins to Settings → System.
+The quiet rules are counted at the end.
 
 ---
 
@@ -135,4 +252,5 @@ Defined in `steward/models/alerts.py`:
 - **`alert_rules`** — one row per configured rule
 - **`alert_states`** — one row per rule, tracks current state and consecutive failure count
 - **`alert_events`** — append-only log of all state transitions and notification outcomes
-- **`plugin_metrics`** — all metric values written by any monitor or plugin
+- **`plugin_metrics`** — all metric values written by any monitor or plugin. Its series index carries the value (`ix_plugin_metrics_series_value`), so a day of one metric is read from the index alone even when nothing is cached; Steward builds it in the background after startup (`core/metric_index.py`), never during a migration
+- **`plugin_metric_latest`** — the newest sample of each (source, resource, metric). Pages that show "what it is now" read this instead of searching history; retention drops a row once its sample leaves the raw window

docs/core/ansible.md

@@ -31,10 +31,9 @@ ansible.sources:
     url: "https://github.com/example/infra.git"
     branch: "main"
     pull_interval_seconds: 300
-    cache_path: "/data/playbook_cache/infra-repo"
 ```
 
-Git sources register a `ScheduledTask` named `ansible_git_pull_<name>` that runs `git pull` on the configured interval.
+Each git source is checked out under `/data/ansible/sources/<name>` (override with `STEWARD_ANSIBLE_SOURCE_DIR`), on the persistent volume, so a redeploy keeps it. One scheduled task, `ansible_git_pulls`, ticks every 30 s and pulls each source once its `pull_interval_seconds` has passed; a source added in Settings is pulled within a tick, with no restart. A clone or pull that runs past 120 s is cut (its process group terminated), a clone that fails is removed so the next attempt starts clean, and git never prompts for credentials.
 
 ---
 
@@ -102,12 +101,20 @@ If a client connects after the run has already completed, the endpoint immediate
 
 | Status | Description |
 |---|---|
+| `queued` | Waiting for a host another run holds, or for a run slot |
 | `running` | Execution in progress |
 | `success` | Playbook exited 0 |
 | `failed` | Playbook exited non-zero |
 | `interrupted` | App restarted while run was in progress |
+| `cancelled` | Stopped by an operator, or by Steward shutting down mid-run |
 
-On startup, the app marks any runs still in `running` state as `interrupted` (see `_mark_interrupted_runs()` in `app.py`).
+On startup, the app marks any runs still `running` or `queued` as `interrupted` (see `_mark_interrupted_runs()` in `app.py`).
+
+### Overlapping runs
+
+- **One run per host.** Before it takes a run slot, a run claims every host it touches. Hosts come from the inventory Steward writes, keyed by address. A run on a source's own inventory file claims that file, because Steward doesn't read its hosts. While another run holds one of its hosts, the run shows `queued` and says which run it waits for. After an hour of waiting it fails and says so.
+- **One run at a time per schedule.** A schedule that falls due while its previous run is still `queued` or `running` skips that firing. The Schedules page shows "Skipped: the previous run was still going" beside that run. The next firing comes an interval later.
+- **One action run at a time per alert rule.** When a rule fires again while the action run from an earlier firing is still going, no new run starts. The firing's event says it was skipped and names the run.
 
 Output stored in the DB is capped at 1 MB. If truncated, `[output truncated]` is appended to the DB column, but the live SSE stream continues unaffected.
 

docs/core/backups.md

@@ -0,0 +1,117 @@
+# Backups: Steward backing itself up, and coming back
+
+Steward sends a backup of itself to an S3 bucket once a day. Fabled Archives
+is the intended target. A backup is one file, and `steward restore` brings
+Steward back from it into an empty database.
+
+## What a backup holds
+
+`steward-<UTC time>.tar`, containing:
+
+| Member | What it is |
+|---|---|
+| `steward.dump` | `pg_dump --format=custom` of Steward's database |
+| `secret.key` | The key Steward's stored passwords and tokens are sealed with |
+| `manifest.json` | Format, Steward version, migration heads, time, and the tables whose rows were left out |
+
+**Kept:**
+- configuration: settings, users, API tokens, hosts, monitors, alert rules, maintenance windows, Ansible inventory and schedules, dashboards, SNMP devices, Prometheus targets, UniFi client names;
+- the audit log;
+- alert history.
+
+**Left out** (the tables come back empty and fill again as Steward runs):
+- metrics and their rollups;
+- monitor check results;
+- Docker, Traefik and UniFi history;
+- Ansible run records;
+- the email queue;
+- password-reset links.
+
+The list is `steward/backup/contents.py`. A test fails when a new table
+appears in neither list, so whoever adds a table decides.
+
+**The secret key is in the file.** That makes a restore complete on its own.
+It also means anyone holding a copy can read every password and key Steward
+keeps. Keep the bucket private.
+
+## Setting it up with Fabled Archives
+
+1. In Archives, make a backup target for Steward with an S3 key. The default
+   append-only key is enough: Steward only uploads and never deletes.
+   How long backups are kept is set on the Archives side.
+2. In Steward, open **Settings → Backups** and fill in:
+   - **Endpoint**: the S3 address Archives shows for the target, without the
+     bucket (for example `http://archives:8081`). Requests are path-style:
+     `endpoint/bucket/file`.
+   - **Bucket**: the target's bucket name.
+   - **Access key** and **Secret key**: from Archives. The secret is sealed at
+     rest; a blank box on save keeps the stored one.
+3. Press **Test the bucket**. A key that may not list ("AccessDenied") still
+   counts as reached: an append-only key can't list, and uploads work anyway.
+4. Tick **Back up every day**, pick the hour (on your clock), and save.
+   **Back up now** makes one straight away.
+
+A day missed while Steward was down is made up when it comes back. A failed
+backup is tried again an hour later, never sooner. The last 30 runs are
+listed on the page, and each one is in the audit log (`backup.made` /
+`backup.failed`).
+
+### Alert when backups stop
+
+While backups are on, Steward records two metrics under the source
+**Steward itself**, resource `backup`:
+
+| Metric | Meaning |
+|---|---|
+| `backup_age_secs` ("Newest backup's age") | Seconds since the last backup was sent, updated every 15 minutes |
+| `backup_bytes` ("Backup size") | The size of the last backup sent |
+
+A rule such as `backup_age_secs > 93600` (26 hours) fires when a daily backup
+has been missed.
+
+## Making a backup by hand
+
+```sh
+docker exec -u app <steward container> steward backup --to /tmp/backups
+```
+
+This writes the file and sends it nowhere. It needs the same database access
+as Steward itself.
+
+## Restoring
+
+`steward restore` writes into an **empty** database. Pass `--wipe` to drop
+everything in the database's public schema first.
+
+1. Stop Steward (scale its service to 0) so nothing writes during the restore.
+2. Download the backup file from Archives.
+3. Run the restore with the image, pointed at the database:
+
+   ```sh
+   docker run --rm \
+     -e STEWARD_DATABASE_URL=postgresql+asyncpg://steward:…@db:5432/steward \
+     -v steward_data:/data -v "$PWD":/restore \
+     git.fabledsword.com/bvandeusen/steward:latest \
+     steward restore /restore/steward-20261012T030000Z.tar
+   ```
+
+   - The backup's secret key is written to `/data/secret.key`.
+   - If `STEWARD_SECRET_KEY` is set instead, it must equal the backup's key,
+     or the restore stops and says so. Read the key with
+     `tar -xOf <file> secret.key`.
+4. Start Steward. On start it runs any migrations the backup is behind on.
+
+**How the restore works.** pg_restore runs in its three sections: schema,
+then data, then constraints and indexes. Between the data and the
+constraints, a kept row that points at a left-out row is put right by its
+foreign key's ON DELETE:
+- an alert event's link to an Ansible run is cleared;
+- a row whose foreign key would cascade or block is removed.
+
+pg_restore 17 writes each section as a script for psql. It runs against
+Postgres 16 with one setting line dropped (`transaction_timeout`, which 16
+doesn't know).
+
+CI checks the whole path on every build. The built image backs up its own
+database, restores into a second one, checks a setting came back and the
+metrics table is empty, then starts Steward on it.

docs/core/mcp.md

@@ -0,0 +1,134 @@
+# MCP server
+
+Steward serves the [Model Context Protocol](https://modelcontextprotocol.io) at
+`/mcp`. An AI client connected to it can look into everything Steward watches
+(hosts, monitors, alerts, metrics, containers, playbook runs) and, with a
+high enough role, change it. A token's role sets that ceiling.
+
+**Source:** `steward/mcp/`. The tools live in `steward/mcp/tools/`, and a
+plugin's tools live in `plugins/<name>/mcp_tools.py`.
+**Transport:** streamable HTTP, stateless. Every request carries its own
+token, so a client survives a Steward restart without reconnecting.
+
+---
+
+## Connecting
+
+1. In Steward, open **AI access** and create a token. Pick the lowest role
+   that does the job.
+2. Copy the token (it is shown once) and use one of the recipes on that
+   page. They carry the real endpoint URL and the new token. For Claude Code:
+
+   ```sh
+   claude mcp add --transport http steward https://steward.example/mcp \
+     --header "Authorization: Bearer stw_…"
+   ```
+
+3. Ask the client to call `whoami`. It answers with the user, the effective
+   role and what that role allows.
+
+Claude Desktop connects through `mcp-remote`. Open WebUI and other clients
+need an MCP server with the **Streamable HTTP** transport and bearer
+authentication. The AI access page shows the recipes for both.
+
+### Behind a reverse proxy
+
+- The URL on the AI access page is **Settings → General → Public base URL**
+  plus `/mcp`. If that setting is empty, the URL is built from the request's
+  host instead.
+- The proxy must pass the `Authorization` header through unchanged.
+- It must not buffer responses. Replies may arrive as server-sent events
+  (`text/event-stream`). For nginx, set `proxy_buffering off;` on the
+  location.
+- Steward does no Host-header (DNS-rebinding) check on `/mcp`, because the
+  bearer token is the security boundary. Serve it over HTTPS.
+
+---
+
+## Security model
+
+| Rule | How it is enforced |
+|---|---|
+| Bearer tokens only, no OAuth | `steward/core/api_tokens.py`. Only a hash is stored, and the token is shown once. |
+| A token never acts above its user | Effective role = min(token role, user's **current** role). Lowering a user lowers every token they hold; disabling the user stops them all. |
+| Default deny | A tool is reachable only if it was registered with a role (`register_tool(fn, role=..., read_only=...)`). `tools/list` shows a caller only the tools their role allows. A call above the role is refused, and the refusal names the role it needs. |
+| Plugin tools follow the plugin | A disabled plugin's tools are hidden and refused. |
+| Dry run first | `run_playbook`, `create_schedule` and `set_alert_action` default to Ansible check mode (`--check --diff`). Changes are applied only when the caller passes `check=false`. A playbook its author flagged `# steward:confirm: true` also needs `confirmed=true`. `prune_docker` previews unless `check=false`. |
+| Secrets stay put | Playbook variables Steward detects as secret reach the run but are never stored, audited or returned. Schedules and alert actions refuse them, since they would have to store them. Inventory variables are listed by name only, and HTTP monitor header values are redacted. |
+| Every change is audited | Each write records an audit event as `<user> via MCP`, with the token id in its detail. `list_audit_events(user="<user>")` includes those events. |
+| Unknown arguments are refused | A misspelt argument is rejected with a "did you mean", rather than silently dropped. |
+
+Tokens never mint or revoke tokens; that happens only on the AI access
+page, in a browser session.
+
+---
+
+## Tools by role
+
+Each role includes the tools of the roles above it in this list.
+
+### Viewer: read everything
+
+| Area | Tools |
+|---|---|
+| Start here | `whoami`, `get_overview` (what is wrong now, ranked, each finding with the call that shows its evidence) |
+| Hosts and metrics | `list_hosts`, `get_host`, `get_host_metrics`, `query_metrics` |
+| Monitors | `list_monitors`, `get_monitor` |
+| Alerts | `list_alerts`, `get_alert_history`, `list_alert_rules`, `get_alert_rule_options`, `list_maintenance_windows` |
+| Metric anomalies | `list_anomalies` (metrics that left their normal for the hour, to review first) |
+| Automation | `list_playbooks`, `get_playbook`, `list_inventory`, `list_ansible_runs`, `get_ansible_run`, `wait_for_run`, `list_schedules` |
+| Audit | `list_audit_events` |
+| Docker plugin | `list_containers`, `get_container`, `list_swarm_services`, `search_container_logs`, `get_docker_disk_usage` |
+| Host agent plugin | `list_agents` |
+
+### Operator: run playbooks, change monitoring
+
+| Area | Tools |
+|---|---|
+| Automation | `run_playbook`, `cancel_ansible_run`, `create_schedule`, `update_schedule`, `delete_schedule`, `update_group_members` (put hosts into an inventory group or take them out; a group alert rule follows within a minute) |
+| Monitors | `create_monitor`, `update_monitor`, `set_monitor_enabled` |
+| Alerts | `create_alert_rule`, `update_alert_rule`, `acknowledge_alert`, `send_test_alert`, `create_maintenance_window`, `end_maintenance_window` |
+| Metric anomalies | `review_anomaly` (mark one normal, so its hours are learned, or real) |
+
+### Admin: destructive and remediation
+
+`set_alert_action` (the playbook a rule runs when it fires), `delete_monitor`,
+`delete_alert_rule`, `prune_docker`.
+
+The client sees each tool's full contract: arguments, what an empty value
+means, and what comes back. That contract is the tool's docstring, so read the
+docstrings in `steward/mcp/tools/` for the detail.
+
+---
+
+## Response conventions
+
+Every tool follows these conventions (`steward/mcp/tools/_shape.py`):
+
+- **Times** are ISO-8601 UTC with a `Z`. Anything with a "when" also carries
+  `age_s`, so the client never does clock arithmetic.
+- **Lists are bounded.** A list response carries `total` and `truncated`. When
+  it was cut, `more` names the argument that fetches the rest. Nothing is cut
+  silently.
+- **Time series** come back as a summary (min/avg/max/p95/last) plus a series
+  of at most 60 points, never raw rows.
+- **Lookup by name or id.** A miss suggests close matches and names the tool
+  that lists the valid choices.
+- **Refusals are readable.** Bad input becomes a tool error that says what was
+  wrong, why, and what to do instead, and nothing is changed. The tools share
+  their validation with the web forms (`monitors/service.py`,
+  `alerts/service.py`), so a value the UI refuses is refused here for the same
+  reason.
+- **Database outages** return a plain "Steward cannot reach its database"
+  rather than a stack trace.
+
+---
+
+## Adding tools
+
+Core tools: add an `async def` to a module in `steward/mcp/tools/` and list it
+in that module's `register()` with its minimum role and `read_only`. A plugin
+exposes `register_mcp_tools(register)`, and optionally `mcp_overview(db)` to
+contribute findings to `get_overview`. See
+[Writing a plugin](../plugins/writing-a-plugin.md). A tool that changes
+anything calls `audit_mcp(...)` from `steward/mcp/tools/_audit.py`.

docs/plugins/host-agent-design.md

@@ -58,9 +58,13 @@ Give Steward a fleet-glance view of remote Linux hosts — current resource usag
 ### File layout on the target host
 
 ```
-/usr/local/lib/steward-agent/agent.py       # the script, target ~300 lines
+/usr/local/lib/steward-agent/agent.py       # the agent (runs as steward-agent)
+/usr/local/lib/steward-agent/maint.py       # the maintenance helper (runs as root, #5766)
 /etc/steward-agent.conf                      # key=value config, 0640 root:steward-agent
-/etc/systemd/system/steward-agent.service    # unit file
+/etc/systemd/system/steward-agent.service    # the agent's unit
+/etc/systemd/system/steward-maint.{path,service}  # starts the helper on a request
+/var/lib/steward-agent/requests/             # the one place the agent writes (0700 steward-agent)
+/var/lib/steward-maint/results/              # the helper's outcomes (root, readable)
                                                 # dedicated system user: steward-agent
 ```
 
@@ -74,8 +78,15 @@ token = a1b2c3d4...
 interval_seconds = 30
 hostname = myhost              # optional; defaults to uname -n
 mounts = /, /mnt/data          # optional; defaults to all real mounts (excluding tmpfs/devtmpfs/etc.)
+docker_logs_enabled = true     # optional; on by default — per-host container-log kill-switch
+docker_log_exclude = watchtower, noisy-svc   # optional; container names whose logs this host skips collecting
 ```
 
+Container-log collection is on by default (no config needed). The two keys above
+are the per-host opt-outs; the operator-facing global toggle + exclude live in
+Settings and are enforced server-side (the push model has no channel to command
+an agent), so they don't require touching a host's conf.
+
 ### Agent internals (function list, not classes)
 
 - `read_config(path)` — parses the conf file into a dict.
@@ -85,10 +96,17 @@ mounts = /, /mnt/data          # optional; defaults to all real mounts (excludin
 - `collect_load()` — reads `/proc/loadavg`, returns `[1m, 5m, 15m]`.
 - `collect_uptime()` — reads `/proc/uptime`, returns seconds since boot (int).
 - `collect_metadata()` — `os.uname()` for kernel + arch, `/etc/os-release` for distro. Called once at startup and cached.
+- `collect_docker_logs(socket, containers, state, exclude)` — (m79) per running
+  container, fetches new log lines over the Docker socket with an incremental
+  `since` cursor kept per container in `state` (a container's first interval
+  seeds from a short tail). Demuxes the multiplexed stream, parses each line's
+  RFC3339 timestamp, caps the whole batch at a byte limit (a marker line records
+  a truncation; the deferred lines come next interval). Folded into the sample as
+  `docker_logs`; omitted when empty.
 - `build_payload()` — assembles a snapshot from all collectors into one dict.
 - `post_payload(url, token, payloads)` — POSTs a list of samples, returns success/failure.
 - `RingBuffer(maxlen=20)` — tiny FIFO wrapper, drops oldest when full.
-- `main_loop()` — the 30s loop: collect → try POST → on failure push to buffer + backoff → on success flush buffer.
+- `main_loop()` — the 30s loop: collect → try POST → on failure push to buffer + backoff → on success flush buffer. Container logs are stripped from a sample before it's buffered (metrics survive an outage; stale logs are dropped).
 
 **Target: ~300 lines total including docstrings.** More than that is a smell that the agent is over-scoping.
 
@@ -99,9 +117,199 @@ Python 3.8+ stdlib only. `urllib.request` for the POST, plus `json`, `os`, `shut
 ### Signals and lifecycle
 
 - **SIGHUP** → re-read config file without restart (lets admins change interval/mounts cleanly).
-- **SIGTERM** (from systemd) → finish current poll cycle → flush buffer one last time → exit.
+- **SIGTERM** (from systemd) → finish current poll cycle → flush buffer one last time, with a `stopping` event saying whether the host is going down too (`systemctl is-system-running` says `stopping`) → exit.
 - **Systemd restart policy** → `Restart=on-failure`, `RestartSec=10`.
 
+### Mounts that don't answer (1.14.0, #5760)
+
+Each mount is read in a daemon thread under one 5 s deadline. A hard-mounted NFS share whose server stops blocks `statvfs` in the kernel with no timeout; read inline, that froze the whole agent. A mount that misses the deadline is sent as `{"mount", "stalled": true}`, and isn't probed again while its last probe is still blocked. Steward counts them as `mounts_stalled`.
+
+### Every Docker read is bounded (1.15.0, #5764)
+
+One Docker answer must arrive within 15 s and fit in 32 MiB, and one log answer within 5 s and 4 MiB. Past either limit the read is cut off, not waited on. A per-read timeout alone never ends a body that keeps trickling in. Container logs are read since each container's cursor, but only its newest 2,000 lines, all containers within 10 s. A container that writes more than that gets a `_steward` note saying older lines were skipped. A container whose answer is too big skips ahead to now. Before this, a 1.13.0 agent re-read a growing backlog every interval and went silent for a day at 2.1 GB. Each sample also carries the agent's own `agent` footprint (`rss_bytes`, `cpu_pct`), which Steward records as `agent_rss_bytes` and `agent_cpu_pct`.
+
+### What the host has waiting (1.16.0, #5765)
+
+Once an hour, and when the agent starts, a background thread asks the package manager what it would upgrade, so a slow apt never holds up a sample. On apt hosts it uses `/usr/lib/update-notifier/apt-check` (pending;security), or `apt list --upgradable` when that's missing. On dnf hosts it uses `dnf -C check-update`, `dnf -C updateinfo list --security`, and `needs-restarting -r` for whether to reboot. None of this needs root or the network: the commands read lists the system's own timers already fetched. Each command has 120 s. The last answer rides every sample as `updates`. Every sample also reads Debian's `/var/run/reboot-required` (and `.pkgs`) as `reboot`. Steward records `updates_pending`, `updates_security` and `reboot_required` (1/0) for rules, and keeps the newest on the registration. The host page's Agent card leads with "Reboot required since …", then "N updates waiting (M security)" or "Up to date", with when the check ran. `state["updates_due"] = True` asks for a check on the next sample; a maintenance job uses it once it has updated the host.
+
+
+### A root helper for updates and reboots (1.17.0, #5766)
+
+The agent can't update or reboot its host: it runs as `steward-agent` under
+`NoNewPrivileges` and `ProtectSystem=strict`, so no sudo is possible, and that
+is how it should stay. A separate helper, `maint.py`, runs as root, started by
+systemd when the agent leaves it a request:
+
+- The agent writes `<id>.json` = `{"id", "action", ["node"]}` into
+  `/var/lib/steward-agent/requests`, the one path its unit lets it write
+  (`ReadWritePaths=-…`; the `-` keeps an old host without the directory
+  starting). It writes a dot-file first and renames it, so the helper never
+  sees half a request.
+- `steward-maint.path` (`PathExistsGlob=…/requests/*.json`) starts
+  `steward-maint.service`, a oneshot running `maint.py` as root.
+- The helper opens each request with `O_NOFOLLOW`, wants a regular file under
+  4 KiB, and accepts only `update`, `reboot`, `drain` or `undrain`, an id of
+  letters, digits and dashes, and a node id in the swarm's own 25-character
+  form. It deletes every request, valid or not, and writes the outcome to
+  `/var/lib/steward-maint/results/<id>.json` (root's; kept a week). A request
+  whose id already has a result is not run again.
+- **update**: `apt-get update` then `dist-upgrade` keeping existing config
+  files, with `NEEDRESTART_MODE=l` so needrestart only lists services rather
+  than restarting the Docker daemon under its containers; or
+  `dnf -y upgrade --refresh`. The result says how many packages changed and
+  whether a reboot is now required.
+- **reboot**: writes `rebooting`, syncs, `systemctl reboot`. Any request after
+  it waits for the next boot.
+- **drain / undrain**: `docker node update --availability drain|active`, on
+  the given node or this host's own. Only a manager can; a worker's helper says
+  so in its result.
+
+The agent reports each result's new state once, as a `maintenance` event with
+its next post that gets through (results older than a day are not news). When
+an update finishes it re-checks what's waiting straight away rather than at
+the next hourly check. Its metadata says `maint_helper: true` once the path
+unit is installed and it can write its request directory; the host's Agent
+panel shows that under Manage agent. **Update agent** installs the helper on a
+host that has an older agent.
+
+So a stolen agent token can make its host update, reboot or drain, and nothing
+else. Steward handing a host a job and following it through the reboot is
+#5767 (below); the cadence that does it one host at a time is #5768.
+
+### A maintenance job, followed through the reboot (1.18.0, #5767)
+
+Steward keeps one `host_maintenance_jobs` row per job and moves it on only from
+what the agents post, so a restart of Steward, or a reboot of the host Steward
+runs on, loses nothing. `plugins/host_agent/maintenance.py` has the states and
+transitions (model-free, unit-tested) and the database side.
+
+```
+queued → updating → [draining →] rebooting → [undraining →] done
+                  ↘ done (no reboot needed, or the job may never reboot)
+any state → failed (a step's result said so) | timed_out (its deadline passed)
+```
+
+- **Hand-off.** Steward's answer to an ingest post carries
+  `"maintenance": [{"id", "action", ["node"]}]`, the step that host's agent
+  should carry out now. It is repeated on every post until a result for that
+  step id arrives; the agent leaves each id for the helper once
+  (`take_maintenance`), and the helper refuses an id it already has a result
+  for, so a repeat, or a re-hand after the agent restarts, never runs twice.
+  A step id is `<job id>-<action>`.
+- **Results** are the `maintenance` events of #5766, matched to the job by step
+  id, so a drain reported by a manager reaches the job of the node it drained.
+- **The reboot completes** when the host's agent sends `started` with a
+  `boot_at` after the reboot was handed out (60 s of clock slack). An agent
+  restart on a host that stayed up has an old `boot_at` and changes nothing.
+- **Reboot only when required**: the update's result says whether a reboot is
+  needed; a job made with "never reboot" stops after the update.
+- **Swarm nodes** are drained first and made active after. The node is found
+  by hostname in the Docker plugin's swarm nodes; a manager drains itself, a
+  worker is drained by a manager whose agent has the helper. With no such
+  manager the job is refused rather than rebooting an undrained node.
+- **Deadlines** (`DEADLINES`): queued 1 h, updating 90 min, draining 10 min,
+  rebooting 30 min, undraining 10 min. A minute tick (`host_agent_maintenance`)
+  times out a job past its deadline.
+- **A maintenance window** scoped to the host's agent metrics
+  (`host_agent/<host>`) opens when the job starts, lasts as long as the rest of
+  the job could take, and ends when it does.
+- Ingest applies all of this in a savepoint of its own: a job going wrong is
+  logged and never costs the host its samples.
+
+On the host's Agent panel: a running job leads what the host has waiting, with
+since-when; a finished one says what it did ("Done: 12 packages updated,
+rebooted"). Under Manage agent, **Update now** (with "never reboot") starts a
+job, or the panel says why one can't start.
+
+### A maintenance cadence: a group, one host at a time (#5768)
+
+`plugins/host_agent/cadence.py`, driven by the same minute tick as the jobs.
+A **plan** (`host_maintenance_plans`) names an inventory group, a weekday, a
+time in a time zone (the creator's, by default), a window and a reboot policy:
+`if_required` (the default), `always` or `never`. The page is **Hosts →
+Maintenance** (`/plugins/host_agent/maintenance`).
+
+- **When.** A plan's slot is the latest weekday-and-time at or before now, in
+  its zone. A slot fires once (`last_slot_at`) and only while its window is
+  open, so a Steward that was down over the slot starts late, not a day late.
+  A new or resumed plan starts from the next slot. **Run now** starts a run
+  without touching the schedule (the Ansible ruling on Run now).
+- **A run** (`host_maintenance_runs`) fixes its hosts at the start, by name:
+  the group's hosts through their linked Ansible targets. A host with no agent,
+  or no maintenance helper, is skipped and listed.
+- **One at a time.** Each host gets a #5767 job (`run_id` on the job). The next
+  host starts only when the last is **back in service**: its agent posting in
+  the last 3 min and, for a swarm node, the node `ready`/`active` and every
+  replicated service at its replicas, judged only from swarm reports made
+  after the job finished. A host not back within 30 min fails the run, and
+  its job is marked failed with the reason.
+- **Stops.** A failed or timed-out job fails the run. No host is started after
+  the window; the run then stops and says how many weren't reached.
+- **Alerting** is an ordinary alert rule on `host_agent/<host>/maintenance_failed`:
+  1 when the host's latest job (in 30 days) failed, timed out or didn't come
+  back into service, 0 once a later one succeeds.
+- The page leads with a running run and its hosts, then failed runs from the
+  last week, then each plan with its schedule, next slot and last five runs.
+  Each host's Agent panel shows its last maintenance job.
+
+### The swarm's health, alertable (1.19.0, #5788)
+
+Every manager's agent reports the whole swarm. From each report the docker plugin (`plugins/docker/swarm_health.py`) records three swarm-wide figures on resource `swarm`, source `docker`:
+
+- `swarm_services_short`: services running fewer tasks than they want. A service scaled to 0 is never short.
+- `swarm_nodes_down`: nodes not `ready`.
+- `swarm_managers_unreachable`: managers the swarm reports as not `reachable`. Agent 1.19.0 sends `ManagerStatus.Reachability`; before that, a manager whose status isn't `ready` counts.
+
+Drained nodes are left out of the last two. A node is drained on purpose (VDNT-Util01 always, any node during its maintenance), so its going away is expected, and its host's own "Agent offline" rule still covers it.
+
+The figures are recorded at most once every 30 s, whichever manager sent them, so three managers make one series and losing one doesn't silence it.
+
+A maintenance job on a swarm node opens a second window over `docker` / `swarm` beside its host's (`maintenance._swarm_window`). An alert the drain or reboot causes then waits for the job, and one still firing when the window ends is sent then.
+
+### Drive health (1.20.0, #5789; 1.20.1, #5806)
+
+Reading SMART needs root, so the root helper does it, as the one action on its fixed list that only reads: `drives`. Hourly, and at start, the agent leaves a `drives` request. The helper runs `smartctl --scan-open` to find the drives itself (the request names none), reads each with `smartctl --json -a -d <type> <dev>`, and writes a summary to `/var/lib/steward-maint/drives.json` (0644). It writes no job result, because a drive read is not a maintenance job. The agent's newest copy rides every sample, as updates do.
+
+smartctl ≥ 7.0 answers in JSON, the same on every distribution and for SATA, SAS and NVMe drives. A missing or older smartctl is reported (`smartctl not installed`, `… too old`) and its text output is never parsed. The helper reports figures only:
+
+- `passed` and `temp_c` for every drive;
+- `critical_warning`, `life_used_pct`, `media_errors` and spare % against threshold for NVMe;
+- reallocated / pending / offline-uncorrectable sectors for ATA, and life used where the drive reports it;
+- grown defects for SCSI.
+
+A drive with no SMART (a virtual disk, a bridge that hides it) reads `available: false`, and one smartctl couldn't open carries `error`. Disks behind a RAID controller share one path (`/dev/bus/0`) and differ by the slot in their type (`megaraid,3`), so the slot is part of the name: `bus0-megaraid3` (1.20.1, #5806). Steward also gives a repeated name an ordinal (`0#2`), so two drives never share a series.
+
+Steward judges the figures in `drives.py`, so the judgment can change without an agent update:
+
+- **failing**: the drive's self-assessment fails, an NVMe critical warning is raised, or spare blocks fall under threshold;
+- **worn**: at least 90% of its rated life is used;
+- **watch**: it has bad sectors or media errors but otherwise passes. Old disks carry a few for years, so a count alone isn't a failure.
+
+The figures become metrics on `<host>:drive:<device>`: `drive_failing`, `drive_life_used_pct`, `drive_bad_sectors` and `drive_temp_c`. The host gets `drives_failing` and `drive_life_used_max_pct` for covering rules; they are recorded only when at least one drive was read, so a VM's virtual disks don't claim "0 failing". The registration keeps the newest report (`drive_health`), and the host page lists it under **Drive health**, what needs attention first.
+
+**Installing smartctl on any common distribution.** `files/install-deps.sh` installs smartmontools with whichever package manager the host has: apt (without recommends, which would pull in a mail server), dnf, yum, zypper, pacman or apk. install.sh and the install / provision / update playbooks all fetch and run that one copy. A package that can't be installed is reported and never fails the install.
+
+### What happened to the host (1.14.0, #5756)
+
+A post may carry `events`, said once each, with the next post that gets through:
+
+- `started` — the agent starting, with `boot_at` (from `/proc/uptime`).
+- `stopping` — on SIGTERM, with `host_shutdown`: true, false, or null when systemd can't say.
+- `lost_contact` — `from` the first post that got no answer (or a proxy's 5xx) `to` the post carrying it.
+
+Steward turns them into the host's downtime (`host_downtime`): **reboot** (a stop, then a boot after it), **agent_stop** (the host stayed up), **unclean** (a boot after the agent was last heard that nobody announced: power loss or a crash) and **lost_contact**. An announced stop explains the silence for `reboot_grace_seconds` (Settings → Plugins → host_agent, 10 minutes): `seconds_since_seen` reads 0 until then, and the host's other rules go quiet with it, as when it is offline. Past the grace, agent offline fires. The events live in memory, so a span the agent saw is lost if it restarts mid-outage; the boot time still reads the restart as unclean.
+
+### Which temperature sensors are thermometers (#5782, #5783)
+
+A board's spare thermistor inputs (a Nuvoton's SYSTIN, AUXTIN*) are often wired to nothing, and the kernel reports them anyway, at whatever the open input reads. The host's temperature (`temp_c_max`) leaves a sensor out when:
+
+- **its value is impossible** (at or below 0 °C, or at or above 125 °C; `counts_as_hottest`), or
+- **it doesn't behave like a thermometer.** Once an hour, `sensors.refresh` judges each sensor against the host's CPU (Intel `Package id N`, AMD `Tctl`/`Tdie`/`TccdN`, else `PECI Agent N`), using the last 7 days of hourly averages. It needs at least 72 shared hours and a CPU that moved at least 10 °C. A sensor is ruled out when it **falls as the CPU warms** (Pearson r ≤ −0.5, the shape of an open input reading the chip's own reference) or **holds within 0.5 °C** while the CPU moved. A wired board or drive sensor that barely follows the CPU (r near 0) stays in.
+
+The verdicts are kept in memory and rebuilt at startup, so ingest and the host page read them without a query. The host page lists the sensors left out under "Not counted", each with its reason. A host with no CPU reading, or too little history, falls back to the value rule alone.
+
+`temp_c_max` can drive an alert rule (#5787), e.g. "Host running hot": all hosts, over 85 °C for 10 consecutive checks. So can each sensor's `temp_c`, and the per-interface and per-disk throughput. `tests/core/test_metric_catalog.py` holds every metric ingest writes to the alert catalog, except the few `NOT_ALERTABLE` names in `plugins/host_agent/__init__.py`.
+
 ### Logging
 
 To stderr only (systemd captures to journal). No file logging, no log rotation. Levels:
@@ -143,7 +351,6 @@ plugins/host_agent/
 │       └── host_agent_001_initial.py    # creates host_agent_registrations table
 ├── templates/
 │   ├── detail.html                      # per-host detail page
-│   ├── widget_table.html                # dashboard fleet table widget
 │   ├── widget_history.html              # dashboard history chart widget
 │   ├── settings_list.html               # plugin settings: list of registered hosts
 │   └── install.sh.j2                    # install script template (Jinja)
@@ -192,6 +399,8 @@ One `ScheduledTask` running every 60 seconds that flags `HostAgentRegistration`
 
 **Note:** the existing alerts system can fire on "no metric for N seconds" which is arguably a cleaner signal. The scheduled task may end up redundant once alert rules exist. Build it anyway as a minimal safety net; it's one query plus one update.
 
+**As built:** the task records `host_agent/<host>/seconds_since_seen` each minute, and "Agent offline" is an ordinary alert rule on it (`> 180`). It counts from the later of `last_seen_at` and the agent's newest post to *arrive* (kept in memory, `scheduler.arrivals`). `last_seen_at` is written in the ingest's own transaction, so a post Steward received but couldn't store (its storage stalled, #5514) would otherwise read as the agent's silence. Nothing counts from before Steward last started watching without a break (#5441).
+
 ### `METRIC_CATALOG` registration
 
 Add to `steward/alerts/routes.py`:
@@ -255,6 +464,7 @@ Content-Type: application/json
 - **`metadata` is sent on every POST**, not just on change. Server-side diff detects actual changes and only writes on change. Cost per POST is one dict — negligible. Benefit: server can cleanly detect agent restarts.
 - **Raw bytes, not percentages, for memory and storage.** Percentages are derived server-side. Changing the "what counts as used" math doesn't require re-releasing the agent.
 - **CPU is the one exception** — reported as a percentage because it's inherently a derivative (delta over time), not a snapshot. The agent must sample twice to compute it.
+- **Container logs (m79) ride in the same push** as `docker_logs`: a list of `{container, stream, ts, line}` records — incremental since the previous interval. This pushes the *same direction* as metrics, so batched log history needs no inbound channel (only sub-second live-follow would). The server ingests them into `docker_logs`, enforces the global toggle + exclude list on ingest, and bounds storage with a per-container size+age ring. `docker_logs` is omitted when there's nothing new.
 
 ### Server expansion into `PluginMetric` rows
 
@@ -324,92 +534,17 @@ sudo sh install.sh
 
 ### Rendered script structure (Jinja template `install.sh.j2`)
 
-```sh
-#!/bin/sh
-# Steward host agent installer
-# Generated for: {{ host_name }}  ({{ host_address }})
-# Steward URL: {{ url }}
-set -e
-
-STEWARD_URL="{{ url }}"
-AGENT_TOKEN="{{ token }}"
-AGENT_VERSION="{{ agent_version }}"
-
-AGENT_USER="steward-agent"
-AGENT_DIR="/usr/local/lib/steward-agent"
-CONF_FILE="/etc/steward-agent.conf"
-UNIT_FILE="/etc/systemd/system/steward-agent.service"
-
-# ── preflight ────────────────────────────────────────────────────────────────
-[ "$(id -u)" = "0" ] || { echo "must run as root (use sudo)"; exit 1; }
-command -v systemctl >/dev/null 2>&1 || { echo "systemd not found — unsupported init system"; exit 1; }
-command -v python3   >/dev/null 2>&1 || { echo "python3 not found — install python3 first"; exit 1; }
-
-# Handle --uninstall
-if [ "${1:-}" = "--uninstall" ]; then
-    systemctl disable --now steward-agent.service 2>/dev/null || true
-    rm -f "$UNIT_FILE" "$CONF_FILE"
-    rm -rf "$AGENT_DIR"
-    systemctl daemon-reload
-    # keep the system user — removing it risks orphaned files elsewhere
-    echo "uninstalled"
-    exit 0
-fi
-
-# ── create system user ───────────────────────────────────────────────────────
-if ! id "$AGENT_USER" >/dev/null 2>&1; then
-    useradd --system --no-create-home --shell /usr/sbin/nologin "$AGENT_USER"
-fi
-
-# ── drop the agent file ──────────────────────────────────────────────────────
-mkdir -p "$AGENT_DIR"
-curl -sSL "${STEWARD_URL}/plugins/host_agent/agent.py" -o "$AGENT_DIR/agent.py"
-chmod 0755 "$AGENT_DIR/agent.py"
-chown root:root "$AGENT_DIR/agent.py"
-
-# ── write config ─────────────────────────────────────────────────────────────
-cat > "$CONF_FILE" <<EOF
-url = $STEWARD_URL
-token = $AGENT_TOKEN
-interval_seconds = 30
-EOF
-chown "root:$AGENT_USER" "$CONF_FILE"
-chmod 0640 "$CONF_FILE"
-
-# ── write systemd unit ───────────────────────────────────────────────────────
-cat > "$UNIT_FILE" <<EOF
-[Unit]
-Description=Steward host agent
-After=network-online.target
-Wants=network-online.target
-
-[Service]
-Type=simple
-User=$AGENT_USER
-ExecStart=/usr/bin/python3 $AGENT_DIR/agent.py
-Restart=on-failure
-RestartSec=10
-NoNewPrivileges=yes
-ProtectSystem=strict
-ProtectHome=yes
-PrivateTmp=yes
-ReadOnlyPaths=/proc /sys
-
-[Install]
-WantedBy=multi-user.target
-EOF
-
-systemctl daemon-reload
-systemctl enable --now steward-agent.service
-
-echo
-echo "Steward host agent $AGENT_VERSION installed and running."
-echo "Check status:  systemctl status steward-agent"
-echo "Logs:          journalctl -u steward-agent -f"
-```
+`plugins/host_agent/templates/install.sh.j2` is the script; it is not copied
+here, because a copy drifts. It checks it runs as root with systemd and
+python3, handles `--uninstall`, creates the `steward-agent` user, fetches
+`agent.py`, writes the config, then fetches the maintenance helper and the
+three units from `/plugins/host_agent/files/` (the same files the playbooks
+install), creates the maintenance directories, and enables
+`steward-maint.path` and the agent.
 
 ### Design points
 
+- **One copy of every file a host installs.** The units and the maintenance helper live in `plugins/host_agent/files/` and are served at `/plugins/host_agent/files/<name>` (a fixed list; anything else is 404). install.sh and the install, provision and update playbooks all fetch them, so the unit can't differ between a host installed by hand and one installed by Ansible (#5766).
 - **Agent binary served by Steward itself.** `GET /plugins/host_agent/agent.py` serves the script bundled with the plugin, so version drift between install-script and agent is impossible — re-running the installer picks up whatever agent the currently-running Steward ships.
 - **Systemd hardening is cheap and correct.** `NoNewPrivileges`, `ProtectSystem=strict`, `ProtectHome`, `PrivateTmp`, `ReadOnlyPaths=/proc /sys`. The agent only needs read access to `/proc`, `/sys`, and mount points; denying everything else narrows blast radius.
 - **Uninstall is a first-class flag**, not a separate script. Same one-liner with `--uninstall` appended.
@@ -423,7 +558,7 @@ echo "Logs:          journalctl -u steward-agent -f"
 
 ### Dashboard widgets (`steward/core/widgets.py`)
 
-- **`host_resources`** — table widget. One row per monitored host: name, CPU %, mem %, disk % (worst mount), load 1m, "last seen Xs ago" with red/yellow/green coloring. Fleet glance.
+- The fleet glance (CPU, mem, disk /, load 1m, stale mark) is the core **Hosts — Overview** widget (`hosts_overview`); the old `host_resources` table was folded into it (#5213).
 - **`host_resource_history`** — chart widget for one host. CPU / mem / disk over a selectable time range (1h, 6h, 24h, 7d). Same pattern as every other history widget — Chart.js.
 
 ### Per-host detail page

docs/plugins/index.yaml.example

@@ -1,113 +0,0 @@
-# steward-plugins / index.yaml
-#
-# This file is the catalog index for the steward plugin repository.
-# It is fetched by the app's Settings → Plugins → Plugin Catalog UI.
-#
-# Steward reads this file from:
-#   https://git.fabledsword.com/bvandeusen/Steward-plugins/raw/branch/main/index.yaml
-#
-# After adding or updating a plugin entry, commit and push — the change is
-# live immediately for anyone whose app fetches the catalog (cache TTL: 5 min).
-#
-# REQUIRED fields for each plugin entry:
-#   name            - must match the plugin's directory name exactly
-#   version         - semver string
-#   download_url    - direct URL to a zip file containing the plugin
-#   checksum_sha256 - SHA-256 hex digest of the zip (leave empty to skip verification)
-#
-# OPTIONAL fields:
-#   description, author, license, min_app_version,
-#   repository_url, homepage, tags
-#
-# Generating a checksum:
-#   sha256sum traefik.zip
-#   shasum -a 256 traefik.zip   (macOS)
-#
-# Download URL conventions:
-#   Gitea release assets (recommended):
-#     https://git.fabledsword.com/bvandeusen/Steward-plugins/releases/download/traefik-v1.0.0/traefik.zip
-#   Upload the zip as a release attachment in Gitea; paste the URL here.
-#   Source archive tarballs work too but release assets are preferred (smaller, plugin-only).
-
-version: 1
-updated: "2026-03-22"
-
-plugins:
-
-  - name: http
-    version: "1.0.0"
-    description: "Synthetic HTTP endpoint monitoring — status code, response time, content match, TLS expiry"
-    author: "Steward"
-    license: "MIT"
-    min_app_version: "0.1.0"
-    repository_url: "https://git.fabledsword.com/bvandeusen/Steward-plugins"
-    homepage: "https://git.fabledsword.com/bvandeusen/Steward-plugins/src/branch/main/http"
-    download_url: "https://git.fabledsword.com/bvandeusen/Steward-plugins/releases/download/http-v1.0.0/http.zip"
-    checksum_sha256: ""
-    tags:
-      - monitoring
-      - http
-      - synthetic
-      - uptime
-
-  - name: docker
-    version: "1.0.0"
-    description: "Docker container status, resource usage, and restart tracking via Docker socket"
-    author: "Steward"
-    license: "MIT"
-    min_app_version: "0.1.0"
-    repository_url: "https://git.fabledsword.com/bvandeusen/Steward-plugins"
-    homepage: "https://git.fabledsword.com/bvandeusen/Steward-plugins/src/branch/main/docker"
-    download_url: "https://git.fabledsword.com/bvandeusen/Steward-plugins/releases/download/docker-v1.0.0/docker.zip"
-    checksum_sha256: ""
-    tags:
-      - containers
-      - docker
-      - infrastructure
-
-  - name: traefik
-    version: "1.0.0"
-    description: "Traefik reverse proxy metrics and access log integration"
-    author: "Steward"
-    license: "MIT"
-    min_app_version: "0.1.0"
-    repository_url: "https://git.fabledsword.com/bvandeusen/Steward-plugins"
-    homepage: "https://git.fabledsword.com/bvandeusen/Steward-plugins/src/branch/main/traefik"
-    download_url: "https://git.fabledsword.com/bvandeusen/Steward-plugins/releases/download/traefik-v1.0.0/traefik.zip"
-    checksum_sha256: ""   # fill in after running: sha256sum traefik.zip
-    tags:
-      - proxy
-      - metrics
-      - access-log
-
-  - name: unifi
-    version: "1.0.0"
-    description: "UniFi Network controller integration — WAN health, devices, clients, DPI"
-    author: "Steward"
-    license: "MIT"
-    min_app_version: "0.1.0"
-    repository_url: "https://git.fabledsword.com/bvandeusen/Steward-plugins"
-    homepage: "https://git.fabledsword.com/bvandeusen/Steward-plugins/src/branch/main/unifi"
-    download_url: "https://git.fabledsword.com/bvandeusen/Steward-plugins/releases/download/unifi-v1.0.0/unifi.zip"
-    checksum_sha256: ""
-    tags:
-      - network
-      - unifi
-      - ubiquiti
-
-  - name: host_agent
-    version: "1.0.0"
-    description: "Remote Linux host resource monitoring via a lightweight Python push agent"
-    author: "Steward"
-    license: "MIT"
-    min_app_version: "0.1.0"
-    repository_url: "https://git.fabledsword.com/bvandeusen/Steward-plugins"
-    homepage: "https://git.fabledsword.com/bvandeusen/Steward-plugins/src/branch/main/host_agent"
-    download_url: "https://git.fabledsword.com/bvandeusen/Steward-plugins/releases/download/host_agent-v1.0.0/host_agent.zip"
-    checksum_sha256: ""
-    tags:
-      - host
-      - monitoring
-      - cpu
-      - memory
-      - storage

docs/plugins/prometheus-widgets.md

@@ -0,0 +1,210 @@
+# Prometheus widget templates
+
+The Prometheus dashboard widget draws one scrape target through a **template**: a short YAML file that says which of the target's series to show and how. Steward ships with one, **Fabled Archives**. You can write your own, paste in one someone shared, and share yours as a file.
+
+A template is data, not code. It names series and lays them out in **blocks**, and Steward draws them. It cannot add a colour, a chart or a formula. Every widget it draws has the same shape:
+
+1. **A verdict:** "All well.", or the worst thing and how many need a look ("2 need a look. Replication: FullNightly-Restic is running, 11h 53m overdue.").
+2. **Each block in turn**, its problems first, saying why, then the healthy rest. After six healthy rows a block says "+N more healthy".
+3. **The target's name and when it was scraped**, linking to its page.
+
+This page covers where templates live, a walk through a first template, each block with an example, the format reference, and the Fabled Archives template.
+
+---
+
+## Where templates live
+
+**Prometheus → Widget templates** lists every template: the built-in ones, and your own. Open one to see its YAML beside a **preview**, which draws it over a target you pick from the values Steward has already stored. The preview redraws as you type and lists each problem with its line number.
+
+| You want to | Do this |
+|---|---|
+| Start from scratch | **New template**, then write or paste YAML. |
+| Use one someone sent you | **New template**, then paste it or **upload** the `.yaml` file. Look at the preview, then **Save template**. |
+| Change a built-in one | Open it, then **Duplicate**. Built-in templates are read-only and change with Steward, so you edit the copy. |
+| Share yours | **Export** downloads the YAML as a file. |
+| Use it on a dashboard | Edit the dashboard, **Add widget → Prometheus**, and pick the target and the template. |
+
+A template saves only when it has no problems, and its `name` must differ from every other template's. A widget remembers which template it draws with, so renaming a template keeps its widgets. Deleting one leaves its widgets saying the template is gone. Your templates are in Steward's backups, and adding, changing or deleting one is recorded in the audit log.
+
+Steward never fetches a template from anywhere. Templates arrive only by being pasted or uploaded.
+
+---
+
+## Your first template
+
+Look at the target's page first (**Prometheus → the target**). It lists every series Steward stored, with its labels and newest value. A template names those series exactly as listed there.
+
+```yaml
+name: My service
+version: 1
+expects: myservice_
+blocks:
+  - values:
+      title: Health
+      items:
+        - label: Queue depth
+          metric: myservice_queue_depth
+          unit: count
+          warn_above: 100
+        - label: Disk free
+          metric: myservice_disk_free_bytes
+          of: myservice_disk_size_bytes
+          warn_below: 10%
+          crit_below: 5%
+  - times:
+      items:
+        - label: Last good run
+          metric: myservice_last_success_timestamp_seconds
+          warn_after: 26h
+          crit_after: 2d
+```
+
+- `name` is what the widget is called, and what the template list and the widget picker show.
+- `version` is the template's own version. Raise it when you change a template you have shared, so the people you sent it to can tell copies apart.
+- `expects` is the start of the metric names it reads. If the target reports none, the widget says so ("is this the template for it?") instead of a column of blanks.
+
+**Metric names as Steward stores them.** A gauge keeps its name. A counter is stored as a rate per second: `requests_total` becomes `requests_per_sec`. A histogram keeps only `_sum_per_sec` and `_count_per_sec`. See [How series are stored](prometheus.md#how-series-are-stored).
+
+---
+
+## The blocks
+
+Every block takes an optional `title`. Without one, it gets the default title shown in its heading below.
+
+### `states`: a state set, one row per thing (default title "States")
+
+A state set has one series per possible state, with the current one at 1 and the rest at 0. Archives' `fabledarchives_backup_target_state{target, state}` is one. Each thing it describes becomes a row naming its current state.
+
+```yaml
+- states:
+    title: Backup targets
+    metric: fabledarchives_backup_target_state
+    name: target                 # the label that names each row
+    healthy: [fresh]
+    quiet: [disabled, waiting]   # shown greyed, never a problem
+    warning: [late]              # anything not listed is a problem
+    columns:
+      - {label: Last backup, metric: fabledarchives_backup_target_last_backup_timestamp_seconds}
+      - {label: Stored, metric: fabledarchives_backup_target_stored_bytes}
+```
+
+- A state in `healthy` is fine, in `quiet` is shown greyed, and in `warning` is amber. **A state in none of the lists is red.** So a state the service adds later shows up as a problem rather than slipping by as healthy.
+- A row reporting no current state is a warning.
+- `columns` add the row's other series beside its state. Each column series is matched to its row by having the same labels. Failing that, it is matched by the label `name` points at.
+- **Deadlines:** `warn_past` and `crit_past` each name a `*_timestamp_seconds` metric per row. Once that time has passed, a row in a healthy state reads as a warning, or a problem for `crit_past` ("is running, 11h 53m overdue"). Quiet and failing rows keep their own reading. Use this when a service ranks a state like "running" above lateness, so a run that never ends would otherwise read as healthy.
+
+### `values`: figures with their units (default title "Figures")
+
+```yaml
+- values:
+    title: Storage
+    items:
+      - label: Disk free
+        metric: fabledarchives_data_disk_free_bytes
+        of: fabledarchives_data_disk_size_bytes   # shown as "16.2 TiB of 41.9 TiB", with a bar
+        warn_below: 10%                            # a share of `of`
+        crit_below: 5%
+        sparkline: true                            # the last day as small bars beside it
+      - label: Stored
+        metric: fabledarchives_person_stored_bytes
+        name: person                               # one row per person
+        where: {kind: files}                       # only the series with kind=files
+```
+
+- `warn_above`, `crit_above`, `warn_below` and `crit_below` take a number, or a share like `10%` when the item has `of`. The red thresholds are checked before the amber ones.
+- Without `name`, an item must match exactly one series. If more match, the row says how many and asks you to narrow it with `where` or `name`.
+- A series the target doesn't report reads "—", "not reported", as a warning.
+- The sparkline's bars are scaled to that series' own peak over the last 24 hours.
+
+### `times`: moments and how long ago (default title "Times")
+
+```yaml
+- times:
+    items:
+      - label: Last good run
+        metric: myservice_last_success_timestamp_seconds
+        warn_after: 26h
+        crit_after: 2d
+```
+
+The metric is a `*_timestamp_seconds`. The row reads "3h ago" (exact time on hover) and turns amber or red once that long has passed. A timestamp of 0 reads "never", which is a warning when the item has a threshold.
+
+### `table`: one metric, two labels (default title "Table")
+
+```yaml
+- table:
+    title: Stored per person
+    metric: fabledarchives_person_stored_bytes
+    rows: person
+    columns: kind
+```
+
+Rows and columns are the values of the two labels, in alphabetical order, under one header row. A missing cell reads "—". Tables show figures only; they don't judge them.
+
+### `info`: labels worth reading (default title "About")
+
+```yaml
+- info:
+    title: Running
+    metric: fabledarchives_build_info
+    labels: [version, channel, commit]   # which labels, in this order; all of them when left out
+```
+
+This block is for a `*_info` or `*_build_info` series, where the labels are the point and the value is always 1. It reads "version 2026.10.12.1717 · channel dev · commit 3605d0a".
+
+---
+
+## Reference
+
+**Top level**
+
+| Key | Needed | What it is |
+|---|---|---|
+| `name` | yes | At most 80 characters; unique among your templates and the built-in ones. |
+| `version` | yes | A whole number from 1. |
+| `blocks` | yes | One to 20 blocks. |
+| `description` | | A line shown in the template list. |
+| `expects` | | The prefix of the metric names it reads, like `fabledarchives_`. |
+| `format` | | The template format; `1`, the only one so far. |
+
+**Keys most blocks and items take**
+
+| Key | What it is |
+|---|---|
+| `metric` | A series name as Steward stores it. |
+| `where` | Labels a series must have, like `{kind: files, state: ok}`. |
+| `name` | On `states`, `values` and `times`: the label that names each row (one row per value). |
+| `unit` | On a `values` item, a `states` column or a `table`: overrides the unit read from the name: `bytes`, `seconds` (a span), `percent`, `per_sec`, `count` (thousands separated: 100,462), `timestamp`. |
+
+Without `unit`, the name decides how a value reads: `_bytes` as bytes, `_seconds` as a span, `_timestamp_seconds` as a moment, `_per_sec` as a rate, `_pct` as a percentage. Anything else is a plain number.
+
+**Spans** (`warn_after`, `crit_after`) are a number of seconds or a number with `s`, `m`, `h` or `d`: `90`, `15m`, `26h`, `2d`.
+
+**Limits:** a template is at most 64 KB, with at most 20 blocks and 30 items in a block.
+
+**What the checker refuses,** each with its line and what to change:
+- an unknown key;
+- a key given twice;
+- anchors and aliases (`&name`, `*name`): write the value out;
+- a share threshold (`10%`) without `of`;
+- a state in two of `healthy`, `quiet` and `warning`;
+- `crit_after` earlier than `warn_after`;
+- a word YAML reads as true or false (`on`, `off`, `yes`, `no`) used as a state or label value. Quote it: `'off'`.
+
+---
+
+## The Fabled Archives template, walked through
+
+The built-in template is a worked example of most of the above. **Duplicate** it to read or change it. It was written against the series Archives actually reports. On 14 Oct 2026 that turned up a nightly replication job that had read "running" for two days without a good copy; the deadlines below exist to catch exactly that.
+
+| Block | What it shows | Why it is written this way |
+|---|---|---|
+| Backup targets (`states`) | Each target Archives receives backups from: fresh, late or missing, with its last backup and what it has stored. | `disabled` and `waiting` are quiet: a target you turned off, or one that hasn't sent its first backup, isn't a problem. `late` is a warning; `missing` (not listed) is a problem. |
+| Replication (`states`) | Each replication job, its last good copy and its next run. | Archives ranks `running` above lateness, so `warn_past` and `crit_past` read the job's own late and missing times. A run that never finishes turns amber, then red. |
+| Storage and integrity (`values`) | Disk free of its size (amber under 10%, red under 5%, last day as bars), files checked of files, malware held, failing notification channels. | Held malware and a failing channel each need you to do something, so any count above 0 is a warning. |
+| Stored per person (`table`) | Each person's files, versions and trash. | Figures to read, not to judge. |
+| Running (`info`) | The version, channel and commit. | |
+
+Left out on purpose (the template's header says so too):
+- `fabledarchives_section_up`: a section Archives couldn't read stops reporting its figures, and the rows that read them then say "not reported".
+- The background queue, HTTP requests and audit counters: these suit charts and alert rules better than a glance.

docs/plugins/prometheus.md

@@ -0,0 +1,111 @@
+# Prometheus Plugin
+
+The Prometheus plugin scrapes any service's Prometheus `/metrics` endpoint and stores what you choose to keep in Steward's metrics. Once stored, a series can be charted, read through the MCP `query_metrics` tool, and watched by an alert rule like any other metric.
+
+It is on by default. Targets are added, tested and edited on its page, **Prometheus** in the sidebar (`/plugins/prometheus/`).
+
+---
+
+## Adding a target
+
+| Field | What it is |
+|---|---|
+| Name | How the target's series are named in Steward (rule pickers, charts). No `/`, `:` or comma. |
+| Metrics address | The full URL, e.g. `http://archives:8081/metrics`. |
+| Bearer token | Sent as `Authorization: Bearer …`. Stored encrypted and never shown again; leave it blank to keep the saved one. |
+| Scrape every | 10 s to an hour. Default 60 s. |
+| Check its HTTPS certificate | Off only for a self-signed certificate you trust. |
+| Keep / Leave out | Name patterns, one a line, `*` for any run of characters. Empty Keep keeps everything; Leave out wins over Keep. |
+| Series cap | The most series stored from one scrape (default 1000). Past it, the rest are counted and shown on the target's page, not stored. |
+
+**Test** scrapes the address once with what is typed, filters included, and says how many series would be stored and under which names. Nothing is saved.
+
+The endpoint hands over everything; Steward chooses what to keep. Start narrow (`yourservice_*`) and widen it if you find you want more: every stored series is a row every scrape.
+
+---
+
+## How series are stored
+
+| On the endpoint | In Steward |
+|---|---|
+| A gauge or an untyped sample | Stored as it is, under its own name. |
+| A counter (`requests_total`) | Stored as a rate per second, `requests_per_sec`, from the second scrape on. A value lower than the last (the service restarted) is skipped rather than read as a negative rate. |
+| A histogram or summary | Its `_sum` and `_count` are stored as rates (`x_sum_per_sec`, `x_count_per_sec`). Buckets and quantiles are not stored. |
+| NaN or ±Inf | Not stored. |
+
+A series without labels is stored under the target's name. A labelled one is stored under `<target>:<label>=<value>,…`, in the order the endpoint gives the labels, for example `Archives:target=Steward,state=late`.
+
+Each target also records four metrics about itself, under its own name:
+
+| Metric | Meaning |
+|---|---|
+| `is_up` | 1 when the last scrape got metrics, 0 when it didn't. |
+| `scrape_ms` | How long the scrape took. |
+| `series_kept` | Series stored from the last scrape. |
+| `series_capped` | Series the last scrape left out because of the cap. |
+
+A value's unit is read from Prometheus naming: `_bytes` as bytes, `_seconds` as a span, `_timestamp_seconds` as a moment ("3h ago"), `_per_sec` as a rate.
+
+A **state set** is one series per possible state, the current one 1 and the rest 0, such as `fabledarchives_backup_target_state{target, state}`. The target page shows it as one row per target (or job), naming the current state; the other states are listed on hover. A metric is read as a state set when its name ends in `_state` and every series has a `state` label reading 0 or 1. The OpenMetrics form, where the label is named after the metric (`foo{foo="a"} 1`), counts too. Each state is still stored as its own series, so an alert rule can watch one state.
+
+**Only what the endpoint reports now is current.** When a scrape succeeds, any series it no longer has loses its newest value. That happens when the endpoint stopped reporting it, for example an old `build_info` after an upgrade, or when a filter or the cap now leaves it out. Its history stays on the chart until retention removes it. A failed scrape leaves the newest values as they were.
+
+Renaming a target moves its stored history to the new name. Deleting one removes its newest values; its history ages out with retention (Settings → Retention).
+
+---
+
+## The dashboard widget
+
+Edit a dashboard, open **Add widget**, and pick **Prometheus** under Integrations. Choose a **target** and a **template**; the widget is named after the template unless you give it a title.
+
+A template says which of the target's series to show and how: backup states, figures with their units, times, a table, the version running. The widget leads with one line naming the worst problem, then each section with its problems first and the healthy rest after. Its **Details** link opens the target's page.
+
+Steward ships with one template, **Fabled Archives**. A target whose series don't match its template says so in the widget rather than showing blanks.
+
+## Widget templates
+
+**Prometheus → Widget templates** lists every template the widget can draw with: the ones built in, and your own. A template of yours that no longer passes the checker (a newer Steward may read the format differently) is listed first with what to change, and the widgets drawn with it say so too.
+
+Open one to see its YAML beside a **preview**: the template drawn over a target you choose, exactly as the widget draws it, from the values Steward has stored. Nothing is fetched from the target to draw it.
+
+- **New template**: write one, or paste in one someone shared. The preview redraws as you type, and lists each problem with its line until there are none.
+- **Upload a file**: on the same page, read a `.yaml` file into the editor to look over before saving.
+- **Duplicate**: copy any template, built-in ones included, under a new name, and change the copy. Built-in templates are read-only: they ship with Steward and change with it.
+- **Export**: download a template's YAML, to keep or to share.
+
+How to write one, every block, and the format reference: [Prometheus widget templates](prometheus-widgets.md).
+
+Saving needs a template that passes the checker and a name no other template has. A widget remembers which template it draws with, so renaming a template keeps its widgets; deleting one leaves its widgets saying the template is gone. Templates are kept in Steward's backups, and adding, changing or deleting one is in the audit log.
+
+## Alerting on a target
+
+In **Alerts → New rule**, choose **Prometheus targets**, then the series (the resource) and the metric. Two rules worth having for any target:
+
+- **It stopped answering:** resource `<target>`, metric `is_up`, `< 1`.
+- **Something was cut off:** resource `<target>`, metric `series_capped`, `> 0`.
+
+---
+
+## Recipe: Fabled Archives
+
+Archives serves `/metrics` on its backup listener (port 8081), which is reachable only on your LAN and the swarm's overlay network, behind a scrape token.
+
+1. In Archives, **Settings → Metrics → Make a token**, named e.g. "Steward". Copy it; it is shown once.
+2. In Steward, **Prometheus → Add target**:
+   - Name `Archives`
+   - Metrics address `http://<archives host>:8081/metrics` (on a shared overlay network, the service name, e.g. `http://archives_app:8081/metrics`)
+   - Bearer token: the one you copied
+   - Keep: `fabledarchives_*`
+3. **Test**, then **Add target**.
+
+Rules worth adding:
+
+| Watch | Resource | Metric | Condition |
+|---|---|---|---|
+| A backup target is late | `Archives:target=<name>,target_id=<id>,state=late` | `fabledarchives_backup_target_state` | `>= 1` |
+| A backup target is missing | `Archives:target=<name>,target_id=<id>,state=missing` | `fabledarchives_backup_target_state` | `>= 1` |
+| A replication job is failing | `Archives:job=<name>,job_id=<id>,state=failing` | `fabledarchives_replication_job_state` | `>= 1` |
+| Archives' data disk is filling | `Archives` | `fabledarchives_data_disk_free_bytes` | `<` your floor |
+| Archives stopped answering | `Archives` | `is_up` | `< 1` |
+
+The target's page lists every stored series with its exact resource name, so copy names from there rather than typing them.

docs/plugins/traefik.md

@@ -12,7 +12,7 @@ On each scrape:
 3. Writes computed metrics to the `traefik_metrics` history table
 4. Calls `record_metric()` for each metric, making them available to alert rules
 
-Request rates and error rates are computed as deltas between the current and previous scrape, divided by elapsed time. Latency percentiles (p50, p95, p99) are approximated via linear interpolation over histogram buckets — these are estimates, not exact percentiles.
+Request rates and error rates are computed as deltas between the current and previous scrape, divided by elapsed time. Latency percentiles (p50, p95, p99) are approximated via linear interpolation over histogram buckets — these are estimates, not exact percentiles. They cover the responses since the previous scrape (each bucket's rise, summed over every status code and method), not every response since Traefik started. A scrape in which a router answered nothing records no latency, shown as "—"; one past the last bucket (5 s by default) reads as that bucket's edge.
 
 ---
 
@@ -79,7 +79,22 @@ The widget auto-refreshes via HTMX polling every `poll_interval` seconds. Fragme
 
 ### Detail Page
 
-The full Traefik page at `/plugins/traefik/` shows all routers with sparkline history charts (last 20 data points) for request rate, p95 latency, and 5xx error rate.
+The full Traefik page at `/plugins/traefik/` reads top to bottom, for the time range picked at the top:
+
+1. **The verdict.** It names the routers that need a look, says why, and says when the problem started ("api, 3.0% of responses are 5xx, started 2h ago"). When a router has been a problem for the whole range, it says so instead. Otherwise it reads "All N routers healthy".
+2. **The request chart.** Requests per second across all routers and the share answered 5xx, the same chart as the dashboard widget.
+3. **A card for each router that needs a look:** why, since when, the 4xx and 5xx status codes it answered in the last 15 minutes, most first ("404 ×1,812, 502 ×40"), and its request rate, bandwidth, p95/p99 latency and 4xx/5xx shares, each with its history. The code is what separates scanners probing for pages (404), a login wall (401/403) and an app that isn't answering (502/503/504). Steward records codes from the version that added them on; until then a card says they aren't recorded yet.
+   Each card then says what its problem usually means and what to do (`plugins/traefik/meaning.py`, #5712):
+   - **Server errors (5xx)** are almost always real. Mostly 500: the app is failing requests. Mostly 502/503/504: Traefik couldn't reach the app, or it timed out.
+   - **Client errors (4xx)** are usually traffic rather than a fault. Mostly 404: scanners and bots. 401/403: a login wall. 429: rate limiting. Mostly 499 is the exception: clients giving up on a slow app.
+   - **Slow responses** compare p95 with p50. When p95 is 4× p50 or more, a few requests are slow (one endpoint or a heavy page). When p50 is close behind, the whole app is slow.
+   - **Next:** a fault links to the swarm service behind the router (its tasks and logs) when Steward can match the name ("codex@swarm" → `codex_app`, "radarr@swarm" → `arr_radarr`), otherwise to Docker → Swarm. Traffic links to Settings → Thresholds.
+4. **Traefik's own figures:** when its config last reloaded, open connections, memory.
+5. **The healthy routers,** folded into one table, then the routers Traefik no longer reports, then TLS certificates by soonest expiry, then Traffic Origin when the access log is on.
+
+A router is graded on its last 15 minutes against Settings → Thresholds. When a problem started is read from the router's history over the range.
+
+Once a router has 5 days of history, its 4xx share and p95 latency are judged against its own normal for that hour (see [Normals and anomalies](../architecture.md)) instead of the fixed cutoffs. One over the cutoff but within its normal is healthy, and the healthy table says it is usual for that router (an error-pages service that answers 4xx all day, a login wall's steady 401s). One out of its normal is a warning that says what is usual, and a metric anomaly to review; the Traefik page lists the routers' anomalies of the last 30 days. How far out counts is set on Settings → Thresholds → Normals (2 points of 4xx share, 50 ms of p95 by default). A 4xx share over fewer requests in the 15 minutes than Settings → Thresholds allows (30 by default) isn't judged at all. The 5xx share is a failure and always keeps its fixed cutoff. Latency is learned only from #5755's per-scrape figures, so its normal arrives 5 days after that release.
 
 ---
 

docs/plugins/writing-a-plugin.md

@@ -28,7 +28,7 @@ license: "MIT"          # any SPDX identifier, e.g. MIT, Apache-2.0, GPL-3.0
 # Optional: prevents loading on older app versions
 min_app_version: "0.1.0"
 
-# Optional: shown in the catalog UI
+# Optional: where the plugin's source lives
 repository_url: "https://github.com/yourname/yourrepo"
 homepage: "https://github.com/yourname/yourrepo/tree/main/myplugin"
 tags:
@@ -323,83 +323,82 @@ Role hierarchy: `admin > operator > viewer`. Requiring `viewer` grants access to
 
 ---
 
-## Publishing to the Catalog
+## Audit Log Entries
 
-The official plugin catalog is hosted at `https://git.fabledsword.com/bvandeusen/Steward-plugins`. Anyone can submit a plugin by opening a pull request — first-party and third-party plugins are treated identically by the catalog system.
+A change made through a plugin's pages is audited with `steward.core.audit.log_audit`, naming the thing it changed by `entity_type` and `entity_id`. So the audit page shows that thing's current name, linked to its page, export `get_audit_entities()` from `__init__.py`:
 
-### Repo layout
-
-```
-steward-plugins/
-├── index.yaml          ← catalog index — the only file Steward fetches
-├── myplugin/
-│   ├── plugin.yaml
-│   ├── __init__.py
-│   └── ...
-└── .github/
-    └── workflows/
-        └── publish.yml ← packages each plugin dir into a zip on release
+```python
+# plugins/myplugin/__init__.py
+def get_audit_entities() -> dict:
+    from .models import Target
+    return {"myplugin_target": (Target, "/plugins/myplugin/targets/{}")}
 ```
 
-### index.yaml entry
+The model needs `id` and `name` columns. An entity deleted since keeps the name the entry recorded in its detail (`detail={"name": ...}`). Add the action's wording to `ACTIONS` in `steward/audit/service.py` so the row reads as a sentence.
 
-Each plugin needs a corresponding entry in `index.yaml`. See `docs/plugins/index.yaml.example` for the full schema. The minimum required fields are:
+---
 
-```yaml
-- name: myplugin
-  version: "1.0.0"
-  description: "What this plugin does"
-  author: "Your Name"
-  license: "MIT"
-  min_app_version: "0.1.0"
-  repository_url: "https://github.com/yourname/yourrepo"
-  homepage: "https://github.com/yourname/yourrepo/tree/main/myplugin"
-  download_url: "https://github.com/yourname/yourrepo/releases/download/myplugin-v1.0.0/myplugin.zip"
-  checksum_sha256: ""   # fill in after zipping — see below
-  tags:
-    - monitoring
+## Agent Tools (MCP)
+
+Steward serves an MCP endpoint at `/mcp` that AI clients reach with a personal API token created on the AI access page. A plugin adds tools to it by exporting an optional `register_mcp_tools(register)` from `__init__.py`:
+
+```python
+# plugins/myplugin/__init__.py
+def register_mcp_tools(register) -> None:
+    from .mcp_tools import list_targets, rescan_target
+    from steward.models.users import UserRole
+    register(list_targets, role=UserRole.viewer, read_only=True)
+    register(rescan_target, role=UserRole.operator, read_only=False)
 ```
 
-### Creating a release zip
+```python
+# plugins/myplugin/mcp_tools.py
+from steward.mcp._context import current_caller, mcp_app
 
-The zip must contain the plugin files at the top level OR inside a single directory named after the plugin. Both layouts work:
 
-```
-# Layout A — flat (preferred)
-myplugin.zip
-├── plugin.yaml
-├── __init__.py
-└── ...
-
-# Layout B — single top-level directory (also accepted, GitHub archive default)
-myplugin.zip
-└── myplugin/
-    ├── plugin.yaml
-    └── ...
+async def list_targets(limit: int = 50) -> dict:
+    """What this tool is for, when to use it over its siblings, what each
+    argument's empty value means, and what comes back. This docstring is the
+    tool's description: it is all the agent knows about it."""
+    app = mcp_app()                 # Quart's current_app does not exist here
+    async with app.db_sessionmaker() as db:
+        ...
+    return {"targets": [...], "total": n, "truncated": n > limit}
 ```
 
-Generate the zip and its checksum:
+- `role` and `read_only` are required. A tool is listed and callable only for callers whose role reaches `role`, and only while the plugin is enabled.
+- Tools run outside Quart's request pipeline: use `mcp_app()` and `current_caller()` from `steward.mcp._context`, never `current_app` or `session`.
+- Refuse by raising `ValueError("what was wrong, and what to do instead")`. The agent reads that text. Any other exception reaches it only as a generic failure.
+- Tool names are global across core and every plugin; a name already taken fails registration and the plugin contributes no tools (logged).
+- Hot-reloading the plugin replaces its tools.
+- Follow the response conventions the core tools use (`steward/mcp/tools/_shape.py`): times via `iso()` plus `age_s()`, lists through `bounded()` so a cut is never silent, series through `summarize()`, and lookups by name or id through `pick()`.
 
-```bash
-cd steward-plugins
-zip -r myplugin.zip myplugin/
-sha256sum myplugin.zip   # paste this into index.yaml checksum_sha256
+### Contributing to get_overview
+
+`get_overview` is the agent's first call: what is wrong right now, ranked. A plugin adds its own problems to it by exporting an optional `mcp_overview(db)` that returns findings built with `steward.mcp.tools.overview.finding`:
+
+```python
+# plugins/myplugin/__init__.py
+async def mcp_overview(db) -> list[dict]:
+    from steward.mcp.tools.overview import finding
+    return [
+        finding("warning", "target_unreachable", t.name, f"{t.name} has not answered for {t.misses} scans",
+                since=t.last_ok_at, ids={"target_id": t.id}, next=f'get_target(target="{t.id}")')
+        for t in await _unreachable_targets(db)
+    ]
 ```
 
-Upload `myplugin.zip` as a GitHub release asset, then update `index.yaml` with the release download URL and checksum.
+- `severity` is `"critical"` (something is down or about to be) or `"warning"` (degraded, or heading that way). Report problems only; a healthy plugin returns `[]`.
+- It is called only while the plugin is enabled, with its own session. If it raises, the overview still returns, and names the plugin under `unavailable`.
 
-### How install works
+---
 
-When a user clicks **Install** in Settings → Plugins:
+## Shipping a Plugin
 
-1. The app downloads the zip from `download_url`
-2. Verifies the SHA-256 checksum (if provided)
-3. Extracts the plugin into its `PLUGIN_DIR`
-4. Runs any pending Alembic migrations for the plugin
-5. Attempts a **hot-reload** — registers the blueprint and scheduled tasks without restarting
-6. If the plugin was previously loaded (blueprint already mounted), a **restart** is required to pick up the new code
+Steward never downloads plugins. There is no catalog and no install button: the app runs with no outbound network.
 
-Hot-reload works reliably for brand-new plugin installs. Updates to already-active plugins require a restart, which can be triggered from the same settings page.
+- **First-party plugins** live in this repository under `plugins/` and ship inside the image, so they version with core.
+- **A third-party plugin** is a folder you put in the external plugin root, `/data/plugins` by default (`STEWARD_PLUGIN_DIR` moves it). Restart Steward and the plugin appears in Settings → Plugins, where you enable it. A bundled plugin of the same name wins over an external one.
 
 ---
 
@@ -413,6 +412,6 @@ Hot-reload works reliably for brand-new plugin installs. Updates to already-acti
 - [ ] Revision IDs prefixed with plugin name
 - [ ] `record_metric()` called inside `session.begin()`
 - [ ] Routes use `@require_role` decorator
+- [ ] MCP tools (if any) registered via `register_mcp_tools(register)` with a role and `read_only`
 - [ ] Templates namespaced under `templates/myplugin/`
 - [ ] Plugin enabled in app settings
-- [ ] `index.yaml` entry added with `download_url` and `checksum_sha256`

docs/reference/code-map.md

@@ -89,6 +89,7 @@ Quick reference for where key functions, models, and entry points live in the co
 | `AlertState` | `steward/models/alerts.py` | `alert_states` |
 | `AlertEvent` | `steward/models/alerts.py` | `alert_events` |
 | `PluginMetric` | `steward/models/metrics.py` | `plugin_metrics` |
+| `PluginMetricLatest` | `steward/models/metrics.py` | `plugin_metric_latest` |
 | `AnsibleRun` | `steward/models/ansible.py` | `ansible_runs` |
 | `User` | `steward/models/users.py` | `users` |
 | `AppSetting` | `steward/models/settings.py` | `app_settings` |
@@ -136,3 +137,25 @@ Quick reference for where key functions, models, and entry points live in the co
 | `steward/migrations/versions/` | Core schema (hosts, users, monitors, alerts, app_settings) |
 | `plugins/traefik/migrations/versions/` | `traefik_metrics` table |
 | `alembic.ini` | Alembic config; `version_locations` lists all migration directories |
+
+---
+
+## MCP server and API tokens
+
+| What | File | Function / Class |
+|---|---|---|
+| Mount `/mcp` (ASGI dispatch, bearer auth, DB gate) | `steward/mcp/server.py` | `mount_mcp(app)` |
+| Server, role registry, refusals | `steward/mcp/server.py` | `StewardMCPServer`, `register_tool()`, `refusal()` |
+| Plugin tools and overview findings | `steward/mcp/server.py` | `register_plugin(name, module)` |
+| Caller and app inside a tool | `steward/mcp/_context.py` | `current_caller()`, `mcp_app()` |
+| Core tool modules | `steward/mcp/tools/` | `register_all()`; one module per area |
+| Response conventions | `steward/mcp/tools/_shape.py` | `iso`, `age_s`, `bounded`, `lookback`, `summarize`, `pick` |
+| Audit a write as "<user> via MCP" | `steward/mcp/tools/_audit.py` | `audit_mcp()` |
+| "What is wrong now" | `steward/mcp/tools/overview.py` | `get_overview()` |
+| Playbooks, runs, schedules, alert actions | `steward/mcp/tools/automation.py` | `run_playbook()`, `wait_for_run()`, … |
+| Monitor validation (forms and tools) | `steward/monitors/service.py` | `monitor_fields()`, `build_config()` |
+| Alert rule / window validation | `steward/alerts/service.py` | `rule_fields()`, `window_fields()`, `acknowledge()` |
+| Docker and host-agent tools | `plugins/docker/mcp_tools.py`, `plugins/host_agent/mcp_tools.py` | `register_tools()`, `mcp_overview()` |
+| Token model | `steward/models/api_tokens.py` | `ApiToken` |
+| Mint, resolve, revoke tokens | `steward/core/api_tokens.py` | `create_token()`, `resolve_token()`, `revoke_token()` |
+| Token management page and connect card | `steward/ai_access/routes.py`, `steward/templates/ai_access/` | `ai_access_bp` |

plugins/docker/__init__.py

@@ -11,7 +11,7 @@ _app: "Quart | None" = None
 def setup(app: "Quart") -> None:
     global _app
     _app = app
-    from .models import DockerContainer, DockerMetric  # noqa: F401 — register with Base
+    from .models import DockerContainer  # noqa: F401 — register with Base
 
     # Publish the per-host persist hook so the host agent can hand off the
     # `docker` array from each sample without importing our models (opportunistic
@@ -21,21 +21,36 @@ def setup(app: "Quart") -> None:
     from steward.core.capabilities import register_capability
     from steward.models.users import UserRole
     from .ingest import persist_host_docker
-    from .retention import run_docker_retention
+    from .retention import prune_docker_logs, run_docker_retention
     register_capability(
         "docker.persist_host_samples", persist_host_docker,
         label="Persist host Docker samples",
         description="Store per-host container state + metrics pushed by the host agent.",
         required_role=UserRole.viewer,
     )
-    # Roll up + prune Docker time-series, driven by the core cleanup task without
-    # it importing our models. Same trusted server-side data-plane role as above.
+    # Prune lifecycle events, driven by the core cleanup task without it
+    # importing our models. Same trusted server-side data-plane role as above.
+    # Container CPU and memory live in the one metric store and its retention.
     register_capability(
         "docker.run_retention", run_docker_retention,
         label="Run Docker retention",
-        description="Roll up old docker_metrics to hourly + prune stale metrics/events.",
+        description="Prune Docker lifecycle events past their window.",
         required_role=UserRole.viewer,
     )
+    register_capability(
+        "docker.prune_logs", prune_docker_logs,
+        label="Prune Docker logs",
+        description="Keep each container's logs within the age window and byte cap.",
+        required_role=UserRole.viewer,
+    )
+    # Each container's CPU (100% is one core) and memory share learn their
+    # normal (#5742), each with its floor from Settings → Thresholds.
+    from steward.core.norms import floor_setting, learn_normal
+    learn_normal("docker", "cpu_pct", floor_setting(app, "container_cpu_pct"))
+    learn_normal("docker", "mem_pct", floor_setting(app, "container_mem_pct"))
+    # A host's agent carries its containers' figures: they go silent with it (#5721).
+    from steward.core.alerts import agent_fed
+    agent_fed("docker", lambda resource: resource.split("/", 1)[0])
 
 
 def get_scheduled_tasks() -> list:
@@ -49,6 +64,29 @@ def get_blueprint():
     return docker_bp
 
 
+def register_mcp_tools(register) -> None:
+    from .mcp_tools import register_tools
+    register_tools(register)
+
+
+async def mcp_overview(db) -> list[dict]:
+    from .mcp_tools import mcp_overview as findings
+    return await findings(db)
+
+
 def get_nav() -> list[dict]:
     # Surfaces in the sidebar's Infrastructure group (fleet-wide container view).
     return [{"label": "Docker", "href": "/plugins/docker/"}]
+
+
+def get_alert_metrics() -> dict[str, list[str]]:
+    """What ingest.py records per container (resource "<host>/<container>"):
+    restart_count is cumulative (alert on crash-looping); is_healthy is 1.0 /
+    0.0 from the container's HEALTHCHECK, only for containers that define one.
+    Per host (resource "<host>"), what Docker keeps on its disk (#5304).
+    Swarm-wide (resource "swarm"), from any manager's report: services short
+    of their replicas, nodes down and managers unreachable, drained nodes
+    aside (#5788, swarm_health.py)."""
+    from .disk_view import DISK_METRICS
+    from .swarm_health import SWARM_METRICS
+    return {"docker": ["cpu_pct", "mem_pct", "restart_count", "is_healthy", *DISK_METRICS, *SWARM_METRICS]}

plugins/docker/disk_view.py

@@ -0,0 +1,82 @@
+# plugins/docker/disk_view.py
+"""What Docker keeps on each host's disk, over time and by image repository
+(#5304–#5306). Model-free, so it is unit-tested without a database.
+
+The agent reports `docker system df` as current state; ingest.py records its
+totals as metrics too (DISK_METRICS, resource = the host's name), so each
+host's growth has a history and an alert rule can watch it.
+"""
+from __future__ import annotations
+
+from datetime import datetime
+
+# metric name → the df summary field it records; "disk_total_bytes" is their sum.
+DISK_PARTS = {
+    "disk_images_bytes": "images_size",
+    "disk_containers_bytes": "containers_size",
+    "disk_volumes_bytes": "volumes_size",
+    "disk_build_cache_bytes": "build_cache_size",
+}
+DISK_METRICS = [*DISK_PARTS, "disk_total_bytes", "disk_reclaimable_bytes"]
+
+
+def disk_metrics(disk: dict) -> dict[str, float]:
+    """The metric samples one df report gives: each part, their total, and
+    what unused images hold."""
+    out = {m: float(int(disk.get(f) or 0)) for m, f in DISK_PARTS.items()}
+    out["disk_total_bytes"] = sum(out.values())
+    out["disk_reclaimable_bytes"] = float(int(disk.get("images_reclaimable") or 0))
+    return out
+
+
+def growth(points: list[tuple[datetime, float]]) -> dict | None:
+    """How much a total changed over the points given (oldest first): the
+    change, the first point's time and value. None with fewer than two."""
+    if len(points) < 2:
+        return None
+    (first_at, first), (_, last) = points[0], points[-1]
+    return {"change": last - first, "since": first_at, "from": first, "to": last,
+            "pct": (last - first) / first * 100 if first else None}
+
+
+def image_repo(repo_tag: str | None) -> tuple[str | None, str | None]:
+    """("ghcr.io/x/app", "1.2") for "ghcr.io/x/app:1.2"; a registry's port is
+    part of the repository ("reg:5000/app" has no tag); a digest reference
+    keeps its digest as the tag; (None, None) for an untagged image."""
+    if not repo_tag or repo_tag == "<none>" or repo_tag.startswith("<none>"):
+        return None, None
+    if "@" in repo_tag:
+        repo, _, digest = repo_tag.partition("@")
+        return repo, digest[:19]
+    head, sep, tag = repo_tag.rpartition(":")
+    if not sep or "/" in tag:
+        return repo_tag, None
+    return head, tag
+
+
+def group_images(images: list[dict]) -> list[dict]:
+    """Images folded under their repository, the most space first.
+
+    Each image: {repo_tag, image_id, size, shared_size, containers}. A group's
+    `own` counts each version's size less the layers it shares with other
+    images, so a repository is not charged twice for one base layer;
+    `reclaimable` is that, for the versions no container uses. Untagged images
+    form one group of their own (repo None)."""
+    groups: dict[str | None, dict] = {}
+    for im in images:
+        repo, tag = image_repo(im.get("repo_tag"))
+        g = groups.setdefault(repo, {"repo": repo, "versions": [], "own": 0, "size": 0,
+                                     "reclaimable": 0, "containers": 0})
+        own = max(int(im.get("size") or 0) - int(im.get("shared_size") or 0), 0)
+        used = int(im.get("containers") or 0)
+        g["versions"].append({"tag": tag, "image_id": im.get("image_id"), "size": int(im.get("size") or 0),
+                              "shared": int(im.get("shared_size") or 0), "own": own, "containers": used})
+        g["own"] += own
+        g["size"] = max(g["size"], int(im.get("size") or 0))
+        g["containers"] += used
+        if not used:
+            g["reclaimable"] += own
+    for g in groups.values():
+        g["versions"].sort(key=lambda v: (v["containers"] == 0, -v["size"]))
+        g["unused"] = sum(1 for v in g["versions"] if not v["containers"])
+    return sorted(groups.values(), key=lambda g: (g["repo"] is None, -g["own"], g["repo"] or ""))

plugins/docker/ingest.py

@@ -34,12 +34,16 @@ def _derive_events(old_state: dict, new_containers: list) -> list:
     """Diff a fresh snapshot against stored per-container state → lifecycle events.
 
     Pure function (no DB) so it's unit-testable. `old_state` maps name → the
-    previously stored {status, health, oom_killed, exit_code}; `new_containers`
+    previously stored {status, health, oom_killed, exit_code, restart_count}; `new_containers`
     is the newest snapshot's list of container dicts. Returns a list of
     (container_name, event, detail) tuples:
 
       start         — a container transitions into running (or a genuinely new
                       container appears already running)
+      die           — also for a new container first seen already exited
+                      non-zero: a task that crashed between two samples (#5049)
+      restart       — Docker restarted it (its restart count went up); said
+                      instead of a start, since that is what happened
       stop          — running → not-running, or a running container is removed
       die           — same as stop but with a non-zero exit code (abnormal)
       oom           — OOMKilled flips False→True
@@ -62,14 +66,22 @@ def _derive_events(old_state: dict, new_containers: list) -> list:
         old = old_state.get(name)
 
         if old is None:
-            # Newly observed container — only a start is meaningful (we have no
-            # prior state to diff a death against).
+            # Newly observed container: a start, or, when it is already down
+            # with a non-zero exit, the death that happened between samples.
             if new_running:
                 events.append((name, "start", c.get("image") or None))
+            elif new_status in _STOPPED_STATES and new_exit not in (None, 0):
+                events.append((name, "die", f"exit {new_exit}"))
             continue
 
         old_running = (old.get("status") or "").lower() == "running"
-        if new_running and not old_running:
+        new_restarts = c.get("restart_count") or 0
+        old_restarts = old.get("restart_count") or 0
+        if new_restarts > old_restarts:
+            gained = new_restarts - old_restarts
+            events.append((name, "restart", f"restart #{new_restarts}" if gained == 1
+                           else f"{gained} restarts, now #{new_restarts}"))
+        if new_running and not old_running and new_restarts <= old_restarts:
             events.append((name, "start", c.get("image") or None))
         elif old_running and not new_running:
             if new_exit not in (None, 0):
@@ -92,6 +104,55 @@ def _derive_events(old_state: dict, new_containers: list) -> list:
     return events
 
 
+def _log_rows(log_batches, host_id: str):
+    """Pure: flatten the agent's log batches → docker_logs row kwargs (no DB).
+
+    `log_batches` is a list of (recorded_at, records) where each record is
+    {"container", "stream", "ts", "line"}. A line's own Docker timestamp wins;
+    recorded_at is the fallback when the agent couldn't parse one. Drops the
+    agent's advisory truncation marker (container "_steward" — it signals a
+    deferral, not a real container), malformed records, and lineless records;
+    normalises an unknown stream to stdout. Unit-testable in isolation.
+    """
+    for recorded_at, records in log_batches:
+        if not isinstance(records, list):
+            continue
+        for rec in records:
+            if not isinstance(rec, dict):
+                continue
+            name = rec.get("container")
+            line = rec.get("line")
+            if not name or name == "_steward" or line is None:
+                continue
+            ts = _parse_started_at(rec.get("ts")) or recorded_at
+            stream = rec.get("stream")
+            if stream not in ("stdout", "stderr"):
+                stream = "stdout"
+            yield {"host_id": host_id, "container_name": str(name)[:255],
+                   "ts": ts, "stream": stream, "line": str(line)}
+
+
+async def _persist_logs(session, host, log_batches) -> None:
+    """Append pushed container log lines (one row per line) for this host.
+
+    Time-series / append-only — the per-container size+age ring (retention)
+    bounds growth, so a chatty container just keeps a shorter window. The global
+    toggle + exclude list (Settings, no restart) are enforced here: this is the
+    authoritative drop point, since the push model has no channel to tell an
+    agent to stop collecting.
+    """
+    from steward.core.settings import get_setting
+    from .models import DockerLog
+
+    if not await get_setting(session, "docker.logs.enabled"):
+        return
+    exclude = set(await get_setting(session, "docker.logs.exclude") or ())
+    for row in _log_rows(log_batches, host.id):
+        if row["container_name"] in exclude:
+            continue
+        session.add(DockerLog(**row))
+
+
 async def _persist_swarm(session, host, swarm: dict) -> None:
     """Upsert this manager's swarm topology; drop rows no longer reported.
 
@@ -120,6 +181,7 @@ async def _persist_swarm(session, host, swarm: dict) -> None:
         row.running = int(s.get("running") or 0)
         row.image = s.get("image")
         row.placement_json = json.dumps(s.get("placement") or [])
+        row.tasks_json = json.dumps(s.get("tasks") or [])
         row.updated_at = now
     stale_services = delete(DockerSwarmService).where(
         DockerSwarmService.host_id == host.id)
@@ -178,6 +240,14 @@ async def _persist_disk(session, host, disk: dict) -> None:
     summary.containers_count = int(disk.get("containers_count") or 0)
     summary.volumes_count = int(disk.get("volumes_count") or 0)
     summary.updated_at = now
+    # The totals as metrics too (#5304): a history of each host's growth, and
+    # something an alert rule can watch. Resource: the host's name.
+    from steward.core.alerts import MetricBatch
+    from .disk_view import disk_metrics
+    batch = MetricBatch()
+    for metric, value in disk_metrics(disk).items():
+        batch.add(source_module="docker", resource_name=host.name, metric_name=metric, value=value)
+    await batch.record(session)
 
     images = disk.get("images") or []
     seen: set[str] = set()
@@ -201,43 +271,116 @@ async def _persist_disk(session, host, disk: dict) -> None:
     await session.execute(stale)
 
 
-async def persist_host_docker(session, host, snapshots, swarm=None, disk=None) -> None:
+async def _persist_status(session, host, status: dict) -> None:
+    """Record whether Docker answered on this host, and why not.
+
+    A host reporting "unavailable" with no row yet never had Docker: nothing to
+    record. Once a host has had Docker, losing it is recorded, so its last
+    containers read as stale rather than as running."""
+    from .models import DockerHostStatus
+
+    row = await session.get(DockerHostStatus, host.id)
+    if row is None:
+        if status["status"] == "unavailable":
+            return
+        row = DockerHostStatus(host_id=host.id)
+        session.add(row)
+    if row.status != status["status"]:
+        row.since = status["at"]
+    row.status = status["status"]
+    row.detail = (status.get("detail") or "")[:255] or None
+    row.reported_at = status["at"]
+
+
+async def _record_swarm_health(session, swarm: dict) -> None:
+    """The swarm-wide figures alert rules watch (#5788), at most once every
+    RECORD_EVERY whichever manager reported."""
+    from datetime import timezone
+    from steward.core.alerts import MetricBatch
+    from . import swarm_health
+    if not swarm_health.due(datetime.now(timezone.utc)):
+        return
+    batch = MetricBatch()
+    for metric, value in swarm_health.figures(swarm).items():
+        batch.add(source_module="docker", resource_name=swarm_health.RESOURCE, metric_name=metric, value=value)
+    await batch.record(session)
+
+
+async def _expire_swarm(session, host) -> None:
+    """Drop the swarm rows a host reported while it was a manager."""
+    from .models import DockerSwarmNode, DockerSwarmService
+
+    await session.execute(delete(DockerSwarmService).where(DockerSwarmService.host_id == host.id))
+    await session.execute(delete(DockerSwarmNode).where(DockerSwarmNode.host_id == host.id))
+
+
+async def persist_host_docker(session, host, snapshots, swarm=None, disk=None,
+                              logs=None, status=None, phases=None) -> None:
     """Upsert containers + time-series + lifecycle events + swarm for one host.
 
     `snapshots` is a list of (recorded_at: datetime, containers: list[dict]) —
     one entry per ingested sample carrying docker data (usually one; more when
-    the agent flushes a backlog). Every snapshot contributes time-series
-    DockerMetric rows; the newest snapshot drives current container state, the
-    alert pipeline, and lifecycle-event derivation. `swarm` is the newest
+    the agent flushes a backlog). Every snapshot's running containers record
+    cpu_pct, mem_pct and mem_usage_bytes into the one metric store, at the
+    snapshot's time; the newest snapshot drives current container state and
+    lifecycle-event derivation. `swarm` is the newest
     sample's swarm object (or None off managers) — persisted when present.
     `disk` is the newest sample's /system/df summary (or None on Docker-less
-    hosts) — persisted when present.
-    """
-    from steward.core.alerts import record_metric
-    from .models import DockerContainer, DockerEvent, DockerMetric
+    hosts) — persisted when present. `logs` is a list of (recorded_at, records)
+    log batches (m79) — appended one row per line when present, independent of
+    whether this sample also carried container metrics. `status` is the newest
+    sample's {"status", "detail", "at", "swarm_manager"} (agent ≥ 1.9.0): an
+    "ok" with an empty snapshot clears the host's containers, and a node that
+    says it is no longer a manager loses the swarm rows it reported as one.
 
+    `phases`, when given, is marked after each part (docker.status, .swarm,
+    .disk, .logs, .metrics, .events, .containers, .alerts, .reap), each
+    flushed before its mark, so a slow ingest names the part (#5181).
+    """
+    from steward.core.alerts import MetricBatch
+    from .models import DockerContainer, DockerEvent
+
+    async def mark(name: str) -> None:
+        if phases is not None:
+            await session.flush()
+            phases.mark(f"docker.{name}")
+
+    if status is not None:
+        await _persist_status(session, host, status)
+        if status.get("swarm_manager") is False:
+            await _expire_swarm(session, host)
+        await mark("status")
     if swarm is not None:
         await _persist_swarm(session, host, swarm)
+        await _record_swarm_health(session, swarm)
+        await mark("swarm")
     if disk is not None:
         await _persist_disk(session, host, disk)
+        await mark("disk")
+    if logs:
+        await _persist_logs(session, host, logs)
+        await mark("logs")
 
     if not snapshots:
         return
     ordered = sorted(snapshots, key=lambda s: s[0])
     latest_at, latest_containers = ordered[-1]
 
-    # Time-series points for every snapshot (running containers with a CPU read).
+    # Time series for every snapshot (running containers with a CPU read), in
+    # the one metric store (#5752): resource "<host>/<container>", so
+    # containers of the same name on different hosts don't collide. Recorded
+    # with the alert batch below, at each snapshot's own time.
+    batch = MetricBatch()
     for recorded_at, containers in ordered:
         for c in containers:
-            if c.get("status") == "running" and c.get("cpu_pct") is not None:
-                session.add(DockerMetric(
-                    host_id=host.id,
-                    container_name=c["name"],
-                    scraped_at=recorded_at,
-                    cpu_pct=c["cpu_pct"],
-                    mem_pct=c.get("mem_pct") or 0.0,
-                    mem_usage_bytes=c.get("mem_usage_bytes") or 0,
-                ))
+            if c.get("status") != "running" or c.get("cpu_pct") is None or not c.get("name"):
+                continue
+            resource = f"{host.name}/{c['name']}"
+            for metric in ("cpu_pct", "mem_pct", "mem_usage_bytes"):
+                if c.get(metric) is not None:
+                    batch.add(source_module="docker", resource_name=resource,
+                              metric_name=metric, value=c[metric], recorded_at=recorded_at)
+    await mark("metrics")
 
     # Snapshot of stored per-container state BEFORE the upsert overwrites it —
     # used to diff lifecycle events. Skip event derivation on the host's first
@@ -247,17 +390,24 @@ async def persist_host_docker(session, host, snapshots, swarm=None, disk=None) -
     )).scalars().all()
     old_state = {
         r.name: {"status": r.status, "health": r.health,
-                 "oom_killed": r.oom_killed, "exit_code": r.exit_code}
+                 "oom_killed": r.oom_killed, "exit_code": r.exit_code,
+                 "restart_count": r.restart_count}
         for r in old_rows
     }
     if old_state:
+        # A stop is dated when Docker says it happened (agent ≥ 1.11), not when
+        # the sample that noticed it was taken.
+        finished = {c.get("name"): _parse_started_at(c.get("finished_at"))
+                    for c in latest_containers}
         for name, event, detail in _derive_events(old_state, latest_containers):
+            at = finished.get(name) if event in ("die", "stop") else None
             session.add(DockerEvent(
                 host_id=host.id, container_name=name,
-                event=event, detail=detail, at=latest_at,
+                event=event, detail=detail, at=at or latest_at,
             ))
+    await mark("events")
 
-    # Current state + alerts from the newest snapshot only.
+    # Current state, and the restart and health alerts, from the newest snapshot only.
     for c in latest_containers:
         name = c.get("name")
         if not name:
@@ -291,23 +441,11 @@ async def persist_host_docker(session, host, snapshots, swarm=None, disk=None) -
         existing.blk_read_bytes = c.get("blk_read_bytes")
         existing.blk_write_bytes = c.get("blk_write_bytes")
 
-        # Alert pipeline — resource is host-scoped so containers of the same name
-        # on different hosts don't collide in the metric/alert namespace.
         resource = f"{host.name}/{name}"
-        if c.get("status") == "running" and c.get("cpu_pct") is not None:
-            await record_metric(
-                session=session, source_module="docker",
-                resource_name=resource, metric_name="cpu_pct", value=c["cpu_pct"],
-            )
-            if c.get("mem_pct") is not None:
-                await record_metric(
-                    session=session, source_module="docker",
-                    resource_name=resource, metric_name="mem_pct", value=c["mem_pct"],
-                )
         # Restart count is alertable regardless of state (crash-looping matters
         # most while the container is down/restarting).
-        await record_metric(
-            session=session, source_module="docker",
+        batch.add(
+            source_module="docker",
             resource_name=resource, metric_name="restart_count",
             value=float(c.get("restart_count", 0) or 0),
         )
@@ -316,12 +454,16 @@ async def persist_host_docker(session, host, snapshots, swarm=None, disk=None) -
         # container).
         health = c.get("health")
         if health in ("healthy", "unhealthy", "starting"):
-            await record_metric(
-                session=session, source_module="docker",
+            batch.add(
+                source_module="docker",
                 resource_name=resource, metric_name="is_healthy",
                 value=1.0 if health == "healthy" else 0.0,
             )
 
+    await mark("containers")
+    await batch.record(session)
+    await mark("alerts")
+
     # Reap containers that vanished from the host's latest listing. Without this,
     # removed one-shot containers (CI job runners, buildkit builders, codex jobs)
     # accumulate forever as permanent "stopped" rows and inflate every count.
@@ -334,3 +476,4 @@ async def persist_host_docker(session, host, snapshots, swarm=None, disk=None) -
     if seen_names:
         reap = reap.where(DockerContainer.name.notin_(seen_names))
     await session.execute(reap)
+    await mark("reap")

plugins/docker/mcp_tools.py

@@ -0,0 +1,703 @@
+"""The docker plugin's MCP tools and its get_overview findings.
+
+Registered through register_mcp_tools() in __init__.py, so they exist only
+while the plugin is enabled. Reads go through service.py, the same functions
+the plugin's pages use.
+"""
+from __future__ import annotations
+
+import json
+from dataclasses import dataclass
+from datetime import datetime, timedelta
+
+from sqlalchemy import select
+
+from steward.ansible.runner import trigger_run
+from steward.ansible.sources import BUILTIN_SOURCE_NAME
+from steward.mcp._context import current_caller, mcp_app
+from steward.mcp.tools._audit import audit_mcp
+from steward.mcp.tools._shape import (
+    age_s, bounded, clamp_limit, db_session, iso, lookback, now_utc, pick, rnd, summarize,
+)
+from steward.mcp.tools.overview import finding
+from steward.models.ansible_inventory import AnsibleTarget
+from steward.models.hosts import Host
+from steward.models.users import UserRole
+
+from . import service as docker_svc
+from .service import PROBLEM_SAYS, PROBLEM_SEVERITY, PROBLEMS, container_problems, problem_since
+from .dedup import dedup_by_container_id
+from .models import DockerContainer, DockerDiskUsage, DockerSwarmNode, DockerSwarmService
+from .swarm_view import build_swarm_services, replaced_tasks, task_time
+
+CRASH_LOOP_DEATHS = docker_svc.CRASH_LOOP_DEATHS
+LOG_LINE_CHARS = 2000
+
+
+# ── pure helpers ─────────────────────────────────────────────────────────────
+
+def cut_middle(text: str, limit: int = LOG_LINE_CHARS) -> str:
+    """Keep both ends of an over-long line and say how much of the middle went."""
+    if text is None or len(text) <= limit:
+        return text
+    keep = limit // 2
+    return f"{text[:keep]} [… {len(text) - 2 * keep} characters omitted …] {text[-keep:]}"
+
+
+def container_row(c, host_names: dict, problems: list[str], now: datetime,
+                  replaced: bool = False, since: datetime | None = None) -> dict:
+    running = (c.status or "").lower() == "running"
+    return {
+        "container_name": c.name,
+        "host_id": c.host_id,
+        "host_name": host_names.get(c.host_id),
+        "image": c.image,
+        "status": c.status,
+        "health": c.health,
+        "problems": problems,
+        # When the worst problem began, from the container's retained events.
+        "problem_since": iso(since),
+        "problem_age_s": age_s(since, now),
+        "exit_code": c.exit_code,
+        "oom_killed": bool(c.oom_killed),
+        "restart_count": c.restart_count,
+        "cpu_pct": rnd(c.cpu_pct),
+        "mem_pct": rnd(c.mem_pct),
+        "mem_usage_bytes": c.mem_usage_bytes,
+        "mem_limit_bytes": c.mem_limit_bytes,
+        "started_at": iso(c.started_at),
+        "uptime_s": age_s(c.started_at, now) if running and c.started_at else None,
+        "compose_project": c.compose_project,
+        "swarm_service": c.service_name,
+        # A swarm task Swarm has replaced: history, judged on its service.
+        "swarm_replaced": replaced if c.service_name else None,
+        "reported_at": iso(c.scraped_at),
+        "reported_age_s": age_s(c.scraped_at, now),
+    }
+
+
+def _from_iso(text: str | None) -> datetime | None:
+    return datetime.fromisoformat(text.replace("Z", "+00:00")) if text else None
+
+
+def container_findings(rows: list[dict]) -> list[dict]:
+    out = []
+    for r in rows:
+        if not r["problems"]:
+            continue
+        worst = r["problems"][0]
+        where = f" on {r['host_name']}" if r["host_name"] else ""
+        detail = f" (exit {r['exit_code']})" if r["exit_code"] not in (None, 0) else ""
+        out.append(finding(
+            PROBLEM_SEVERITY[worst], f"container_{worst}", r["container_name"],
+            f"container {r['container_name']}{where} {' and '.join(PROBLEM_SAYS[p] for p in r['problems'])}{detail}",
+            since=_from_iso(r["problem_since"]),
+            ids={"host_id": r["host_id"], "container_name": r["container_name"]},
+            next=f'get_container(container="{r["container_name"]}", host="{r["host_id"]}")'))
+    return out
+
+
+def task_error_text(err: dict | None) -> str:
+    """Swarm's own account of a task's end, as one clause."""
+    if not err:
+        return ""
+    why = err.get("error") or err.get("message") or err.get("state") or "unknown"
+    code = f", exit {err['exit_code']}" if err.get("exit_code") not in (None, 0) else ""
+    where = f" on {err['node']}" if err.get("node") else ""
+    return f"{why}{code}{where}"
+
+
+# Services that started failing within this long of each other read as one
+# finding: they are usually one node's event (#5722).
+TOGETHER = timedelta(minutes=5)
+
+
+def service_findings(services: list[dict]) -> list[dict]:
+    """Swarm services failing now: crash-looping, or short of replicas, each
+    naming Swarm's last task error. One back to all its replicas is history,
+    kept on the Docker page, not a finding (#5722). Several short of replicas
+    since within TOGETHER of each other are one finding naming them all."""
+    out, short = [], []
+    for s in services:
+        replicas = f"{s['running']} of {s['desired']} replicas running"
+        last = task_error_text(s.get("last_error"))
+        tail = f"; last task error: {last}" if last else ""
+        ids = {"service_name": s["name"]}
+        since = s.get("trouble_since")
+        if s.get("crash_looping"):
+            out.append(finding(
+                "critical", "swarm_service_crash_looping", s["name"],
+                f"swarm service {s['name']} keeps failing: {s['failures_1h']} tasks failed in the "
+                f"last hour, {replicas}{tail}", since=since, ids=ids,
+                next="list_swarm_services()"))
+        elif s["state"] != "healthy":
+            short.append((s, replicas, tail))
+    for group in _together(short):
+        if len(group) == 1:
+            s, replicas, tail = group[0]
+            out.append(finding(
+                "critical" if s["state"] == "down" else "warning", f"swarm_service_{s['state']}",
+                s["name"], f"swarm service {s['name']} has {replicas}{tail}",
+                since=s.get("trouble_since"), ids={"service_name": s["name"]},
+                next="list_swarm_services()"))
+            continue
+        names = [s["name"] for s, _r, _t in group]
+        down = any(s["state"] == "down" for s, _r, _t in group)
+        out.append(finding(
+            "critical" if down else "warning", "swarm_services_short", f"{len(group)} swarm services",
+            f"{len(group)} swarm services started failing together and are short of replicas: "
+            + ", ".join(f"{s['name']} ({r})" for s, r, _t in group),
+            since=min(s["trouble_since"] for s, _r, _t in group),
+            ids={"service_names": names}, next="list_swarm_services()"))
+    return out
+
+
+def _together(short: list[tuple]) -> list[list[tuple]]:
+    """Short services grouped by when they started failing: each group's
+    starts lie within TOGETHER of the one before. One without a start stands alone."""
+    timed = sorted((x for x in short if x[0].get("trouble_since")), key=lambda x: x[0]["trouble_since"])
+    groups: list[list[tuple]] = []
+    for x in timed:
+        if groups and x[0]["trouble_since"] - groups[-1][-1][0]["trouble_since"] <= TOGETHER:
+            groups[-1].append(x)
+        else:
+            groups.append([x])
+    return groups + [[x] for x in short if not x[0].get("trouble_since")]
+
+
+# ── loading ──────────────────────────────────────────────────────────────────
+
+async def _host_names(db) -> dict[str, str]:
+    return dict((await db.execute(select(Host.id, Host.name))).all())
+
+
+def _crash_loop_deaths() -> int:
+    return docker_svc.crash_loop_deaths(mcp_app().config)
+
+
+async def _replaced(db, containers) -> set[tuple]:
+    services = list((await db.execute(select(DockerSwarmService))).scalars())
+    return replaced_tasks(containers, services)
+
+
+async def _rows(db, now: datetime) -> list[dict]:
+    """Every container with its problems. A swarm task Swarm replaced has none:
+    its failure, if it was one, belongs to its service (service_findings)."""
+    host_names = await _host_names(db)
+    return [container_row(r["container"], host_names, r["problems"], now,
+                          replaced=r["replaced"], since=r["since"])
+            for r in await docker_svc.ranked_containers(db, now, _crash_loop_deaths())]
+
+
+async def _swarm(db, now: datetime):
+    containers = dedup_by_container_id(await docker_svc.list_containers(db))
+    services = list((await db.execute(select(DockerSwarmService))).scalars())
+    nodes = list((await db.execute(select(DockerSwarmNode))).scalars())
+    host_names = await _host_names(db)
+    node_hostname = {n.node_id: n.hostname for n in nodes if n.hostname}
+    built = build_swarm_services(containers, services, node_hostname, host_names,
+                                 now=now, crash_loop_deaths=_crash_loop_deaths())
+    return built["services"], nodes, host_names
+
+
+async def mcp_overview(db) -> list[dict]:
+    """get_overview findings: failing containers and under-replicated swarm services."""
+    now = now_utc()
+    rows = await _rows(db, now)
+    services, _, _ = await _swarm(db, now)
+    return container_findings(rows) + service_findings(services)
+
+
+# ── tools ────────────────────────────────────────────────────────────────────
+
+async def list_containers(host: str = "", status: str = "", problem: str = "",
+                          compose_project: str = "", limit: int = 100) -> dict:
+    """Docker containers across every host that runs a Steward agent with Docker.
+
+    host: a host's name or id (list_hosts); empty for all hosts.
+    status: Docker's state — "running", "exited", "restarting", "paused",
+      "created" or "dead"; empty for all.
+    problem: "any" for containers with something wrong, or one of
+      "restarting", "crash_looping" (died repeatedly within the last hour;
+      the count is set under Settings → Thresholds, default 3), "unhealthy"
+      (failing its healthcheck), "oom_killed", "exited_nonzero" (died with an
+      error in the last 24h; exit 137/143 without an OOM kill is a stop, not
+      an error); empty for no filter.
+    compose_project: only containers from this Compose project.
+    limit: default 100, max 500.
+
+    Containers with problems come first. Each row's `problems` lists what is
+    wrong, most severe first, and `problem_since` / `problem_age_s` say when
+    the worst of them began (null when Steward's retained events cannot say).
+    `reported_age_s` is how old the agent's last report is, so a large value
+    means the row may be stale.
+
+    A swarm task's container that Swarm has replaced (a rolling update, or a
+    retry after a failure) has `swarm_replaced: true` and no problems: Swarm
+    restarts a task as a new container, so failures are counted on the
+    service — list_swarm_services shows them.
+    """
+    if problem and problem != "any" and problem not in PROBLEMS:
+        raise ValueError(f"problem {problem!r} is not 'any' or one of {', '.join(PROBLEMS)}.")
+    limit = clamp_limit(limit, 100, 500)
+    now = now_utc()
+    async with db_session() as db:
+        rows = await _rows(db, now)
+        if host:
+            hosts = list((await db.execute(select(Host))).scalars())
+            h = pick(host, hosts, kind="host", list_tool="list_hosts")
+            rows = [r for r in rows if r["host_id"] == h.id]
+    if status:
+        rows = [r for r in rows if (r["status"] or "").lower() == status.lower()]
+    if problem == "any":
+        rows = [r for r in rows if r["problems"]]
+    elif problem:
+        rows = [r for r in rows if problem in r["problems"]]
+    if compose_project:
+        rows = [r for r in rows if r["compose_project"] == compose_project]
+    rows.sort(key=lambda r: (not r["problems"], (r["status"] or "") != "running",
+                             (r["host_name"] or "").lower(), r["container_name"]))
+    return bounded(rows, limit, key="containers",
+                   more="Pass a larger `limit` (max 500), or filter by `host`, `status` or `problem`.")
+
+
+@dataclass
+class RemovedContainer:
+    """A container Docker has removed, known only by its retained logs and events."""
+    host_id: str
+    name: str
+    last_seen: datetime
+
+
+_REMOVED_HINT = ("It may have been removed: Swarm replaces a dead task with a new container under "
+                 "a new name. list_containers(status=\"exited\") lists stopped containers, and "
+                 "search_container_logs(host=...) searches a whole host.")
+
+
+async def _pick_container(db, container: str, host: str, *, removed: bool = False):
+    """The container named `container`: a live one, else (with `removed`) one
+    Docker has removed whose logs or events Steward still holds, returned as a
+    RemovedContainer. Exact names beat partial ones, and live beats removed."""
+    containers = await docker_svc.list_containers(db)
+    host_id = None
+    if host:
+        hosts = list((await db.execute(select(Host))).scalars())
+        host_id = pick(host, hosts, kind="host", list_tool="list_hosts").id
+        containers = [c for c in containers if c.host_id == host_id]
+    want = (container or "").strip()
+    if not want:
+        raise ValueError("pass a container name; list_containers lists them.")
+    live = {(c.host_id, c.name) for c in containers}
+
+    def _one(found, what):
+        if len(found) > 1:
+            raise ValueError(f"{len(found)} hosts have {what} named {want!r}; pass `host` "
+                             f"(one of: {', '.join(sorted(f[0] for f in found))}).")
+        return found[0] if found else None
+
+    same = [c for c in containers if c.name == want] or \
+           [c for c in containers if c.name.lower() == want.lower()]
+    if same:
+        return _one([(c.host_id, c) for c in same], "a container")[1]
+    if removed:
+        gone = [r for r in await docker_svc.retained_names(db, want, exact=True, host_id=host_id)
+                if (r[0], r[1]) not in live]
+        if gone:
+            return RemovedContainer(*_one(gone, "a removed container"))
+    # One more chance: a unique container whose name contains it (swarm task
+    # names look like "stack_svc.1.k2j…").
+    partial = [c for c in containers if want.lower() in c.name.lower()]
+    if len(partial) == 1:
+        return partial[0]
+    gone = []
+    if removed and not partial:
+        gone = [r for r in await docker_svc.retained_names(db, want, exact=False, host_id=host_id)
+                if (r[0], r[1]) not in live]
+        if len(gone) == 1:
+            return RemovedContainer(*gone[0])
+    names = sorted(c.name for c in partial)[:10] + [f"{r[1]} (removed)" for r in gone[:10]]
+    hint = f" Containers whose names contain it: {', '.join(names)}." if names else ""
+    services = sorted({c.service_name for c in partial if c.service_name})
+    if services:
+        hint += (f" Swarm services among them: {', '.join(services[:10])}; search_container_logs "
+                 "takes a service's name and searches every task it has run.")
+    raise ValueError(f"no container named {want!r}{' on that host' if host else ''}.{hint} "
+                     f"{'' if names else _REMOVED_HINT + ' '}list_containers lists them.")
+
+
+def _removed_note(c: RemovedContainer) -> str:
+    return (f"{c.name} no longer exists: Docker removed it (Swarm replaces a failed task with a new "
+            f"container). This is what Steward retained; it was last seen {iso(c.last_seen)}.")
+
+
+async def get_container(container: str, host: str = "", range: str = "24h", events: int = 20) -> dict:
+    """One container in depth: state, resources over time and lifecycle events.
+
+    container: the container's name, as list_containers shows. A unique part of
+      a name also works (swarm task names are long).
+    host: the host's name or id. Needed only when several hosts run a
+      container with this name.
+    range: the window for `cpu_pct_history` and `mem_pct_history` (min, avg,
+      max, p95, last and a ≤60-point series) — "1h", "6h", "24h" (default),
+      "7d" or "30d". `cpu_pct` and `mem_pct` are the latest reported values.
+    events: how many lifecycle events (start, restart, stop, die, oom, health_change)
+      to return, newest first (default 20, max 100).
+
+    A container Docker has since removed (a dead Swarm task, a `docker rm`)
+    still resolves by its exact name, or by a part of it that matches one
+    removed container, while Steward retains its logs or events: the reply
+    has `removed: true`, `last_seen_at`, its history and its events, and no
+    live state. A live container has `removed: false`.
+    """
+    since, key = lookback(range)
+    events = clamp_limit(events, 20, 100)
+    now = now_utc()
+    async with db_session() as db:
+        c = await _pick_container(db, container, host, removed=True)
+        if isinstance(c, RemovedContainer):
+            return await _removed_container(db, c, since, key, events, now)
+        gone = (c.host_id, c.name) in await _replaced(db, await docker_svc.list_containers(db))
+        failures = (await docker_svc.recent_failures_by_container(
+            db, now - timedelta(hours=24))).get((c.host_id, c.name), [])
+        host_names = await _host_names(db)
+        cpu = await docker_svc.container_points(db, c.host_id, c.name, "cpu_pct", since)
+        mem = await docker_svc.container_points(db, c.host_id, c.name, "mem_pct", since)
+        evs = await docker_svc.container_events(db, c.host_id, c.name, events)
+        sick = await docker_svc.unhealthy_since(db, [(c.host_id, c.name)])
+    hour_ago = now - timedelta(hours=1)
+    recent = sum(1 for e in failures if e.at >= hour_ago)
+    try:
+        ports = json.loads(c.ports_json or "[]")
+    except ValueError:
+        ports = []
+    problems = [] if gone else container_problems(c, recent, bool(failures), _crash_loop_deaths())
+    began = problem_since(problems, failures, hour_ago, sick.get((c.host_id, c.name)))
+    return {
+        **container_row(c, host_names, problems, now, replaced=gone, since=began),
+        "removed": False,
+        "container_id": c.container_id,
+        "ports": ports,
+        "net_rx_bytes": c.net_rx_bytes, "net_tx_bytes": c.net_tx_bytes,
+        "blk_read_bytes": c.blk_read_bytes, "blk_write_bytes": c.blk_write_bytes,
+        "range": key,
+        "cpu_pct_history": summarize(cpu),
+        "mem_pct_history": summarize(mem),
+        "events": _event_rows(evs, now),
+        "next": f'search_container_logs(container="{c.name}", host="{c.host_id}")',
+    }
+
+
+def _event_rows(evs, now: datetime) -> list[dict]:
+    return [{"event": e.event, "detail": e.detail, "at": iso(e.at), "age_s": age_s(e.at, now)}
+            for e in evs]
+
+
+async def _removed_container(db, c: RemovedContainer, since, key: str, events: int,
+                             now: datetime) -> dict:
+    """get_container for a container Docker has removed: what Steward kept."""
+    host_names = await _host_names(db)
+    cpu = await docker_svc.container_points(db, c.host_id, c.name, "cpu_pct", since)
+    mem = await docker_svc.container_points(db, c.host_id, c.name, "mem_pct", since)
+    evs = await docker_svc.container_events(db, c.host_id, c.name, events)
+    return {
+        "container_name": c.name, "host_id": c.host_id, "host_name": host_names.get(c.host_id),
+        "removed": True, "note": _removed_note(c),
+        "last_seen_at": iso(c.last_seen), "last_seen_age_s": age_s(c.last_seen, now),
+        "range": key,
+        "cpu_pct_history": summarize(cpu),
+        "mem_pct_history": summarize(mem),
+        "events": _event_rows(evs, now),
+        "next": f'search_container_logs(container="{c.name}", host="{c.host_id}", range="24h")',
+    }
+
+
+async def list_swarm_services(limit: int = 100) -> dict:
+    """Docker Swarm services with their replicas per node, and the swarm's nodes.
+
+    `state` is "healthy" (all desired replicas running), "degraded" (some) or
+    "down" (none). `replicas` are the service's current tasks: each names its
+    host and container status, and one marked `ghost` runs on a node with no
+    Steward agent, so Steward knows only how many tasks are placed there.
+    Tasks Swarm replaced (rolling updates, retries) are left out and counted
+    in `previous_tasks`.
+
+    From Swarm's own task records (a manager's agent 1.8.0 or later;
+    `tasks_known: false` and nulls otherwise):
+    - `failures_1h` / `failures_24h`: tasks that failed or were rejected.
+      Swarm retries a failed task as a new one, so this, not any container's
+      restart count, is how a crash loop shows. Swarm keeps about 5 old tasks
+      per replica, which bounds the count.
+    - `crash_looping`: failures_1h reached the crash-loop count set under
+      Settings → Thresholds (default 3).
+    - `last_error`: the newest task that failed, was rejected, or carries an
+      error while Swarm retries it — `error` is Swarm's own text (e.g. "No
+      such image", "task: non-zero exit (1)"), with `exit_code`, `node`, `at`.
+    - `trouble_since` / `trouble_age_s`: when the current trouble began — a
+      crash loop's first failure in the hour, a short service's first failed
+      task in its failing slots, a recovered service's first failure in 24h.
+      Swarm keeps few old tasks, so a long outage reads as a lower bound.
+
+    Services with a problem come first. `nodes` lists every swarm node once,
+    with its role (manager/worker), availability and status.
+    limit: default 100, max 500.
+    """
+    limit = clamp_limit(limit, 100, 500)
+    now = now_utc()
+    async with db_session() as db:
+        services, nodes, host_names = await _swarm(db, now)
+    seen, node_rows = set(), []
+    for n in sorted(nodes, key=lambda n: n.updated_at, reverse=True):
+        if n.node_id in seen:
+            continue
+        seen.add(n.node_id)
+        node_rows.append({"node_id": n.node_id, "hostname": n.hostname, "role": n.role,
+                          "leader": n.leader, "availability": n.availability, "status": n.status,
+                          "reported_by": host_names.get(n.host_id)})
+    node_rows.sort(key=lambda n: n["hostname"].lower())
+    order = {"down": 0, "degraded": 1, "healthy": 2}
+
+    def _rank(s):
+        return (not s["crash_looping"], order.get(s["state"], 3), not s["failures_24h"], s["name"])
+
+    def _error_row(err):
+        if err is None:
+            return None
+        at = task_time(err)
+        return {"error": err["error"], "message": err["message"], "state": err["state"],
+                "exit_code": err["exit_code"], "node": err["node"],
+                "at": iso(at), "age_s": age_s(at, now) if at else None}
+
+    rows = [
+        {
+            "service_name": s["name"], "mode": s["mode"], "state": s["state"],
+            "running": s["running"], "desired": s["desired"],
+            "crash_looping": s["crash_looping"],
+            "failures_1h": s["failures_1h"], "failures_24h": s["failures_24h"],
+            "last_error": _error_row(s["last_error"]),
+            "trouble_since": iso(s["trouble_since"]),
+            "trouble_age_s": age_s(s["trouble_since"], now),
+            "previous_tasks": s["previous_tasks"], "tasks_known": s["tasks_known"],
+            "replicas": [
+                {"ghost": True, "host": r["host"], "count": r["count"]} if r["ghost"] else
+                {"ghost": False, "host": r["host"], "host_id": r["host_id"], "container_name": r["name"],
+                 "status": r["status"], "health": r["health"], "restart_count": r["restart_count"],
+                 "cpu_pct": rnd(r["cpu_pct"]), "mem_pct": rnd(r["mem_pct"])}
+                for r in s["replicas"]
+            ],
+        }
+        for s in sorted(services, key=_rank)
+    ]
+    return {"nodes": node_rows, **bounded(rows, limit, key="services")}
+
+
+async def search_container_logs(container: str = "", host: str = "", text: str = "",
+                                range: str = "1h", stream: str = "", before: str = "",
+                                context: int = 0, limit: int = 50) -> dict:
+    """Search the container log lines Steward has retained, newest first.
+
+    container: one container's name (as list_containers shows), or a Swarm
+      service's name (as list_swarm_services shows) to search every task
+      container the service has run, live and removed, so a search reaches
+      back across deploys; each match then has the `slot` and `task_id` that
+      wrote it. Empty searches every container (narrow with `host`). A
+      container Docker has removed still works while its lines are retained:
+      the reply then has `container_removed: true` and a `note` saying when it
+      was last seen.
+    host: a host's name or id; empty for all hosts.
+    text: case-insensitive substring to match; empty matches every line.
+    range: how far back — "1h" (default), "6h", "24h", "7d" or "30d". Logs are
+      kept per container up to a size and age limit, so older lines may be gone.
+    stream: "stdout" or "stderr"; empty for both.
+    before: an ISO time; returns only lines older than it. Pass a response's
+      `next_before` to page further back.
+    context: lines of surrounding log (0-5) to include with each match, from
+      the same container. Only with `container` set.
+    limit: matches to return, default 50, max 200.
+
+    Lines longer than 2000 characters keep both ends, and the middle is
+    replaced by a marker saying how much was cut.
+    """
+    since, key = lookback(range)
+    limit = clamp_limit(limit, 50, 200)
+    context = max(0, min(int(context or 0), 5))
+    if context and not container:
+        raise ValueError("context needs `container`: it shows lines around a match within one container's log.")
+    if stream and stream not in ("stdout", "stderr"):
+        raise ValueError(f"stream {stream!r} is not 'stdout' or 'stderr'.")
+    before_dt = None
+    if before:
+        try:
+            before_dt = datetime.fromisoformat(before.replace("Z", "+00:00"))
+        except ValueError as exc:
+            raise ValueError(f"before {before!r} is not an ISO-8601 time; pass a response's next_before.") from exc
+    async with db_session() as db:
+        host_id = name = removed = service = tasks = None
+        if host:
+            hosts = list((await db.execute(select(Host))).scalars())
+            host_id = pick(host, hosts, kind="host", list_tool="list_hosts").id
+        if container:
+            service, tasks = await _as_service(db, container.strip(), host_id)
+            if service is None:
+                c = await _pick_container(db, container, host, removed=True)
+                host_id, name = c.host_id, c.name
+                removed = c if isinstance(c, RemovedContainer) else None
+        rows = await docker_svc.search_logs(db, since=since, host_id=host_id, name=name,
+                                            containers=tasks, text=text, stream=stream,
+                                            before=before_dt, limit=limit + 1)
+        more = len(rows) > limit
+        rows = rows[:limit]
+        host_names = await _host_names(db)
+        matches = []
+        for r in rows:
+            m = {"at": iso(r.ts), "container_name": r.container_name, "host_id": r.host_id,
+                 "host_name": host_names.get(r.host_id), "stream": r.stream, "line": cut_middle(r.line)}
+            if service:
+                m["slot"], m["task_id"] = docker_svc.task_of(service, r.container_name)
+            if context:
+                pre, post = await docker_svc.log_context(db, r.host_id, r.container_name, r.ts, context)
+                m["before"] = [cut_middle(x.line) for x in pre]
+                m["after"] = [cut_middle(x.line) for x in post]
+            matches.append(m)
+    out = {"range": key, "matches": matches, "returned": len(matches), "truncated": more}
+    if service:
+        out["service"] = service
+        out["tasks_searched"] = len(tasks)
+    if removed is not None:
+        out["container_removed"] = True
+        out["note"] = _removed_note(removed)
+        if removed.last_seen < since:
+            out["note"] += " That is before this range; pass a longer `range`."
+    if more:
+        # Full precision: `at` is cut to the second for reading, and a cursor
+        # cut the same way would skip lines logged earlier in that second.
+        out["next_before"] = rows[-1].ts.isoformat()
+        out["more"] = "More matches are older; pass next_before as `before`, or narrow with `text`."
+    return out
+
+
+async def _as_service(db, want: str, host_id: str | None):
+    """(service, its task containers) when `want` names a Swarm service rather
+    than a container (#5468): every task container Steward holds anything for,
+    live and removed, kept to `host_id` when given (none there reads as no
+    lines, not as a missing name). (None, None) when it is not a service, or
+    a container carries that exact name, which keeps precedence."""
+    if not want:
+        return None, None
+    exact = (await db.execute(select(DockerContainer.name).where(
+        DockerContainer.name == want).limit(1))).first()
+    if exact is not None:
+        return None, None
+    tasks = await docker_svc.service_containers(db, want)
+    if not tasks:
+        return None, None
+    return want, [(h, n) for h, n in tasks if host_id is None or h == host_id]
+
+
+async def get_docker_disk_usage(host: str = "") -> dict:
+    """Docker disk usage per host: images, containers, volumes and build cache, in bytes.
+
+    host: a host's name or id; empty for every host. `images_reclaimable_bytes`
+    is what pruning unused images would free. An admin can prune from
+    Steward's Docker → Disk page.
+    """
+    async with db_session() as db:
+        summaries = list((await db.execute(select(DockerDiskUsage))).scalars())
+        host_names = await _host_names(db)
+        if host:
+            hosts = list((await db.execute(select(Host))).scalars())
+            hid = pick(host, hosts, kind="host", list_tool="list_hosts").id
+            summaries = [s for s in summaries if s.host_id == hid]
+    rows = [
+        {
+            "host_id": s.host_id, "host_name": host_names.get(s.host_id),
+            "images_total": s.images_total, "images_active": s.images_active,
+            "images_bytes": s.images_size, "images_reclaimable_bytes": s.images_reclaimable,
+            "containers_count": s.containers_count, "containers_bytes": s.containers_size,
+            "volumes_count": s.volumes_count, "volumes_bytes": s.volumes_size,
+            "build_cache_bytes": s.build_cache_size,
+            "reported_at": iso(s.updated_at), "reported_age_s": age_s(s.updated_at),
+        }
+        for s in summaries
+    ]
+    rows.sort(key=lambda r: -(r["images_reclaimable_bytes"] or 0))
+    return {"hosts": rows, "total": len(rows)}
+
+
+PRUNE_BUILTIN_PLAYBOOK = "maintenance/docker_prune.yml"
+_STOPPED = ("exited", "created", "dead")
+
+
+def prune_preview(what: str, usage, stopped: int) -> dict:
+    """What a prune of `what` would remove, from the host's latest disk report."""
+    out: dict = {}
+    if what in ("containers", "system"):
+        out["stopped_containers"] = stopped
+    if what == "images":
+        out["unused_images_bytes"] = usage.images_reclaimable if usage else None
+    if what == "system":
+        out["dangling_images"] = "yes (unused but tagged images stay)"
+        out["build_cache_bytes"] = usage.build_cache_size if usage else None
+        out["unused_networks"] = "yes"
+    return out
+
+
+async def prune_docker(host: str, what: str, check: bool = True) -> dict:
+    """Reclaim Docker disk on a host. Previews unless check=false.
+
+    host: the Steward host's name or id. It needs a linked Ansible target
+      (list_inventory shows the links); the prune runs through Ansible as
+      Steward's built-in docker_prune playbook.
+    what: "containers" removes stopped containers; "images" removes ALL
+      unused images, not only dangling ones; "system" is `docker system
+      prune -f`: stopped containers, dangling images, unused networks and
+      build cache. Volumes are never pruned.
+    check: true (default) starts nothing and returns `would_remove`, sized
+      from the host's latest Docker disk report. Ansible's own check mode
+      skips the prune commands, so a check-mode run would show nothing.
+      Pass false to run the prune.
+
+    Admin only, as on the Docker → Disk page. A real prune returns `run_id`
+    (follow it with wait_for_run) and is recorded in the audit log.
+    """
+    what = (what or "").strip()
+    if what not in docker_svc.PRUNE_TARGETS:
+        raise ValueError(f"what {what!r} is not one of {', '.join(docker_svc.PRUNE_TARGETS)}.")
+    async with db_session() as db:
+        h = pick(host, (await db.execute(select(Host))).scalars(), kind="host", list_tool="list_hosts")
+        target = (await db.execute(
+            select(AnsibleTarget).where(AnsibleTarget.host_id == h.id))).scalars().first()
+        usage = await db.get(DockerDiskUsage, h.id)
+        stopped = sum(1 for c in await docker_svc.list_containers(db, h.id)
+                      if (c.status or "").lower() in _STOPPED)
+    if target is None:
+        raise ValueError(f"host {h.name!r} has no linked Ansible target, so Steward cannot reach it "
+                         "to prune. Link one under Ansible → Inventory. Nothing was run.")
+    preview = {
+        "host": h.name,
+        "what": what,
+        "ansible_target": target.name,
+        "would_remove": prune_preview(what, usage, stopped),
+        "disk_report_age_s": age_s(usage.updated_at) if usage else None,
+    }
+    if check:
+        return {**preview, "check": True,
+                "summary": f"Nothing was run. Pass check=false to prune {what} on {h.name}."}
+    caller = current_caller()
+    run, _src, err = await trigger_run(
+        mcp_app(), source_name=BUILTIN_SOURCE_NAME, playbook_path=PRUNE_BUILTIN_PLAYBOOK,
+        inventory_scope=f"steward:target:{target.id}",
+        params={"extra_vars_map": docker_svc.prune_extra_vars(what)},
+        triggered_by=caller.user_id, origin="mcp")
+    if err:
+        raise ValueError(f"the prune could not start: {err}.")
+    await audit_mcp("docker.prune.started", "host", h.name,
+                    {"host_id": h.id, "what": what, "run_id": run.id})
+    return {**preview, "check": False, "run_id": run.id,
+            "next": f'wait_for_run(run_id="{run.id}")'}
+
+
+def register_tools(register) -> None:
+    for fn in (list_containers, get_container, list_swarm_services,
+               search_container_logs, get_docker_disk_usage):
+        register(fn, role=UserRole.viewer, read_only=True)
+    register(prune_docker, role=UserRole.admin, read_only=False)

plugins/docker/migrations/env.py

@@ -15,7 +15,7 @@ sys.path.insert(0, str(Path(__file__).parent.parent.parent.parent))
 
 from steward.models.base import Base
 import steward.models  # noqa: F401
-from plugins.docker.models import DockerContainer, DockerMetric  # noqa: F401
+from plugins.docker.models import DockerContainer  # noqa: F401
 
 config = context.config
 if config.config_file_name is not None:

plugins/docker/migrations/versions/docker_009_container_logs.py

@@ -0,0 +1,48 @@
+"""Docker container logs table
+
+Adds docker_logs — one row per container log line, pushed by the host agent and
+folded into its metrics push. Time-series, host-scoped (container names are only
+unique within a host). Bounded by a per-container size+age ring in retention, so
+a chatty container keeps a shorter window rather than growing without limit.
+Additive create_table + twin indexes (viewer lookup by (host, container, ts);
+age-cutoff prune by ts).
+
+Revision ID: docker_009_container_logs
+Revises: docker_008_bigint_mem
+Create Date: 2026-07-19
+"""
+from typing import Sequence, Union
+from alembic import op
+import sqlalchemy as sa
+
+revision: str = "docker_009_container_logs"
+down_revision: Union[str, None] = "docker_008_bigint_mem"
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    op.create_table(
+        "docker_logs",
+        sa.Column("id", sa.String(length=36), nullable=False),
+        sa.Column("host_id", sa.String(length=36), nullable=False),
+        sa.Column("container_name", sa.String(length=255), nullable=False),
+        sa.Column("ts", sa.DateTime(timezone=True), nullable=False),
+        sa.Column("stream", sa.String(length=8), nullable=False, server_default="stdout"),
+        sa.Column("line", sa.Text(), nullable=False, server_default=""),
+        sa.ForeignKeyConstraint(["host_id"], ["hosts.id"], ondelete="CASCADE"),
+        sa.PrimaryKeyConstraint("id"),
+    )
+    op.create_index("ix_docker_logs_host_id", "docker_logs", ["host_id"])
+    op.create_index("ix_docker_logs_container_name", "docker_logs", ["container_name"])
+    op.create_index("ix_docker_logs_host_container_time",
+                    "docker_logs", ["host_id", "container_name", "ts"])
+    op.create_index("ix_docker_logs_ts", "docker_logs", ["ts"])
+
+
+def downgrade() -> None:
+    op.drop_index("ix_docker_logs_ts", table_name="docker_logs")
+    op.drop_index("ix_docker_logs_host_container_time", table_name="docker_logs")
+    op.drop_index("ix_docker_logs_container_name", table_name="docker_logs")
+    op.drop_index("ix_docker_logs_host_id", table_name="docker_logs")
+    op.drop_table("docker_logs")

plugins/docker/migrations/versions/docker_010_swarm_tasks.py

@@ -0,0 +1,30 @@
+"""Docker swarm service tasks column
+
+Adds docker_swarm_services.tasks_json — each service's newest tasks as a
+manager's agent (≥ 1.8.0) reports them: slot, node, desired and current state,
+Swarm's error and message, exit code and time. Lets Steward tell a task Swarm
+replaced from one that crashed, count a service's failures across task
+retries, and say why its last task failed. Additive column; no DROP+recreate.
+
+Revision ID: docker_010_swarm_tasks
+Revises: docker_009_container_logs
+Create Date: 2026-10-06
+"""
+from typing import Sequence, Union
+from alembic import op
+import sqlalchemy as sa
+
+revision: str = "docker_010_swarm_tasks"
+down_revision: Union[str, None] = "docker_009_container_logs"
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    op.add_column("docker_swarm_services",
+                  sa.Column("tasks_json", sa.Text, nullable=False,
+                            server_default="[]"))
+
+
+def downgrade() -> None:
+    op.drop_column("docker_swarm_services", "tasks_json")

plugins/docker/migrations/versions/docker_011_host_status.py

@@ -0,0 +1,36 @@
+"""Docker host status
+
+Adds docker_host_status: per host, whether Docker answered at the agent's
+last report (ok / unavailable / error) and why not. Lets Steward show a host
+whose Docker went away as such, instead of its last containers as running
+(#5152). Additive create_table.
+
+Revision ID: docker_011_host_status
+Revises: docker_010_swarm_tasks
+Create Date: 2026-10-06
+"""
+from typing import Sequence, Union
+from alembic import op
+import sqlalchemy as sa
+
+revision: str = "docker_011_host_status"
+down_revision: Union[str, None] = "docker_010_swarm_tasks"
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    op.create_table(
+        "docker_host_status",
+        sa.Column("host_id", sa.String(length=36), nullable=False),
+        sa.Column("status", sa.String(length=16), nullable=False),
+        sa.Column("detail", sa.String(length=255), nullable=True),
+        sa.Column("since", sa.DateTime(timezone=True), nullable=False),
+        sa.Column("reported_at", sa.DateTime(timezone=True), nullable=False),
+        sa.ForeignKeyConstraint(["host_id"], ["hosts.id"], ondelete="CASCADE"),
+        sa.PrimaryKeyConstraint("host_id"),
+    )
+
+
+def downgrade() -> None:
+    op.drop_table("docker_host_status")

plugins/docker/migrations/versions/docker_012_drop_metrics.py

@@ -0,0 +1,56 @@
+"""Container CPU and memory move to the one metric store
+
+Drops docker_metrics and docker_metrics_hourly (#5752). Container cpu_pct,
+mem_pct and mem_usage_bytes are recorded into plugin_metrics as source
+"docker", resource "<host>/<container>", and roll up with every other metric.
+CPU and memory % were already recorded there for alerting, so their history
+stays; memory in bytes starts fresh.
+
+Revision ID: docker_012_drop_metrics
+Revises: docker_011_host_status
+Create Date: 2026-10-14
+"""
+from typing import Sequence, Union
+from alembic import op
+import sqlalchemy as sa
+
+revision: str = "docker_012_drop_metrics"
+down_revision: Union[str, None] = "docker_011_host_status"
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    op.execute("DROP TABLE IF EXISTS docker_metrics_hourly")
+    op.execute("DROP TABLE IF EXISTS docker_metrics")
+
+
+def downgrade() -> None:
+    op.create_table(
+        "docker_metrics",
+        sa.Column("id", sa.String(36), primary_key=True),
+        sa.Column("host_id", sa.String(36), sa.ForeignKey("hosts.id", ondelete="CASCADE"),
+                  nullable=False, index=True),
+        sa.Column("container_name", sa.String(255), nullable=False, index=True),
+        sa.Column("scraped_at", sa.DateTime(timezone=True), nullable=False, index=True),
+        sa.Column("cpu_pct", sa.Float, nullable=False),
+        sa.Column("mem_pct", sa.Float, nullable=False),
+        sa.Column("mem_usage_bytes", sa.BigInteger, nullable=False),
+    )
+    op.create_index("ix_docker_metrics_host_container_time", "docker_metrics",
+                    ["host_id", "container_name", "scraped_at"])
+    op.create_table(
+        "docker_metrics_hourly",
+        sa.Column("id", sa.String(36), primary_key=True),
+        sa.Column("host_id", sa.String(36), sa.ForeignKey("hosts.id", ondelete="CASCADE"),
+                  nullable=False),
+        sa.Column("container_name", sa.String(255), nullable=False),
+        sa.Column("bucket", sa.DateTime(timezone=True), nullable=False),
+        sa.Column("cpu_pct", sa.Float, nullable=False),
+        sa.Column("mem_pct", sa.Float, nullable=False),
+        sa.Column("mem_usage_bytes", sa.BigInteger, nullable=False),
+        sa.Column("sample_count", sa.Integer, nullable=False),
+        sa.UniqueConstraint("host_id", "container_name", "bucket",
+                            name="uq_docker_metrics_hourly_bucket"),
+    )
+    op.create_index("ix_docker_metrics_hourly_bucket", "docker_metrics_hourly", ["bucket"])

plugins/docker/models.py

@@ -1,13 +1,12 @@
 # plugins/docker/models.py
 from __future__ import annotations
-import uuid
 from datetime import datetime, timezone
 from sqlalchemy import (
     BigInteger, Boolean, DateTime, Float, ForeignKey, Index, Integer, String, Text,
-    UniqueConstraint,
 )
 from sqlalchemy.orm import Mapped, mapped_column
 from steward.models.base import Base
+from steward.core.ids import time_ordered_id
 
 
 class DockerContainer(Base):
@@ -61,69 +60,6 @@ class DockerContainer(Base):
     blk_write_bytes: Mapped[int | None] = mapped_column(BigInteger, nullable=True)
 
 
-class DockerMetric(Base):
-    """Time-series CPU/memory per container — one row per sample per running
-    container, scoped to the reporting host."""
-    __tablename__ = "docker_metrics"
-
-    id: Mapped[str] = mapped_column(
-        String(36), primary_key=True, default=lambda: str(uuid.uuid4())
-    )
-    host_id: Mapped[str] = mapped_column(
-        String(36), ForeignKey("hosts.id", ondelete="CASCADE"), nullable=False, index=True
-    )
-    container_name: Mapped[str] = mapped_column(String(255), nullable=False, index=True)
-    scraped_at: Mapped[datetime] = mapped_column(
-        DateTime(timezone=True), nullable=False,
-        default=lambda: datetime.now(timezone.utc),
-        index=True,
-    )
-    cpu_pct: Mapped[float] = mapped_column(Float, nullable=False, default=0.0)
-    mem_pct: Mapped[float] = mapped_column(Float, nullable=False, default=0.0)
-    # BigInteger: per-container memory usage exceeds 2^31 bytes (~2.1 GB) routinely.
-    mem_usage_bytes: Mapped[int] = mapped_column(BigInteger, nullable=False, default=0)
-
-    # Per-container history lookups filter on (host_id, container_name) then sort
-    # by time — a composite index keeps the rows() sparkline queries cheap.
-    __table_args__ = (
-        Index("ix_docker_metrics_host_container_time",
-              "host_id", "container_name", "scraped_at"),
-    )
-
-
-class DockerMetricHourly(Base):
-    """Hourly rollup of docker_metrics — avg cpu/mem per container per hour.
-
-    Raw per-sample rows (~2880/container/day at 30s) are pruned beyond a short
-    window; before deletion they're aggregated here so multi-day history stays
-    cheap to store and query. One row per (host, container, hour bucket); the
-    unique constraint lets retention upsert idempotently if it re-runs before the
-    raw rows are deleted. `bucket` is the hour-truncated sample time.
-    """
-    __tablename__ = "docker_metrics_hourly"
-
-    id: Mapped[str] = mapped_column(
-        String(36), primary_key=True, default=lambda: str(uuid.uuid4())
-    )
-    host_id: Mapped[str] = mapped_column(
-        String(36), ForeignKey("hosts.id", ondelete="CASCADE"), nullable=False
-    )
-    container_name: Mapped[str] = mapped_column(String(255), nullable=False)
-    bucket: Mapped[datetime] = mapped_column(DateTime(timezone=True), nullable=False)
-    cpu_pct: Mapped[float] = mapped_column(Float, nullable=False, default=0.0)
-    mem_pct: Mapped[float] = mapped_column(Float, nullable=False, default=0.0)
-    mem_usage_bytes: Mapped[int] = mapped_column(BigInteger, nullable=False, default=0)
-    sample_count: Mapped[int] = mapped_column(Integer, nullable=False, default=0)
-
-    __table_args__ = (
-        # One bucket per container per host — the conflict target for the
-        # idempotent rollup upsert; doubles as the history-query index.
-        UniqueConstraint("host_id", "container_name", "bucket",
-                         name="uq_docker_metrics_hourly_bucket"),
-        Index("ix_docker_metrics_hourly_bucket", "bucket"),
-    )
-
-
 class DockerEvent(Base):
     """Lifecycle events derived by diffing consecutive host snapshots.
 
@@ -139,14 +75,14 @@ class DockerEvent(Base):
     __tablename__ = "docker_events"
 
     id: Mapped[str] = mapped_column(
-        String(36), primary_key=True, default=lambda: str(uuid.uuid4())
+        String(36), primary_key=True, default=time_ordered_id
     )
     host_id: Mapped[str] = mapped_column(
         String(36), ForeignKey("hosts.id", ondelete="CASCADE"), nullable=False
     )
     container_name: Mapped[str] = mapped_column(String(255), nullable=False)
     event: Mapped[str] = mapped_column(String(16), nullable=False)
-    # start | stop | die | oom | health_change
+    # start | restart | stop | die | oom | health_change
     detail: Mapped[str | None] = mapped_column(String(255), nullable=True)
     # e.g. "exit 137", "healthy→unhealthy", image on start
     at: Mapped[datetime] = mapped_column(
@@ -163,6 +99,43 @@ class DockerEvent(Base):
     )
 
 
+class DockerLog(Base):
+    """Container log lines pushed by the host agent — one row per line.
+
+    Time-series, scoped to the reporting host (container names are only unique
+    within a host, same identity as docker_containers). The agent tails each running
+    container incrementally and folds new lines into its metrics push; `ts` is the
+    line's own Docker timestamp. Bounded by a per-container size+age ring
+    (retention), so a chatty container just keeps a shorter window rather than
+    growing without limit.
+    """
+    __tablename__ = "docker_logs"
+
+    id: Mapped[str] = mapped_column(
+        String(36), primary_key=True, default=time_ordered_id
+    )
+    host_id: Mapped[str] = mapped_column(
+        String(36), ForeignKey("hosts.id", ondelete="CASCADE"), nullable=False, index=True
+    )
+    container_name: Mapped[str] = mapped_column(String(255), nullable=False, index=True)
+    ts: Mapped[datetime] = mapped_column(
+        DateTime(timezone=True), nullable=False,
+        default=lambda: datetime.now(timezone.utc),
+    )
+    stream: Mapped[str] = mapped_column(String(8), nullable=False, default="stdout")
+    # stdout | stderr
+    line: Mapped[str] = mapped_column(Text, nullable=False, default="")
+
+    # Viewer filters on (host_id, container_name) and sorts by time; the ring
+    # prune walks the same key newest-first. A second index on `ts` alone serves
+    # the age-cutoff delete. Twin-index idiom, mirroring docker_events.
+    __table_args__ = (
+        Index("ix_docker_logs_host_container_time",
+              "host_id", "container_name", "ts"),
+        Index("ix_docker_logs_ts", "ts"),
+    )
+
+
 class DockerSwarmService(Base):
     """A Swarm service as seen by a manager host: desired vs running replicas.
 
@@ -186,6 +159,11 @@ class DockerSwarmService(Base):
     # A manager sees every task, so this captures cross-node placement that the
     # local docker_containers rows (this host only) can't.
     placement_json: Mapped[str] = mapped_column(Text, nullable=False, default="[]")
+    # The service's newest tasks, newest first, as the manager's agent shaped
+    # them: [{"task_id", "slot", "node_id", "desired_state", "state", "message",
+    # "error", "exit_code", "at"}], JSON. Swarm's own record of which tasks it
+    # replaced and why each one stopped; "[]" from agents older than 1.8.0.
+    tasks_json: Mapped[str] = mapped_column(Text, nullable=False, default="[]")
     updated_at: Mapped[datetime] = mapped_column(
         DateTime(timezone=True), nullable=False,
         default=lambda: datetime.now(timezone.utc),
@@ -262,3 +240,24 @@ class DockerSwarmNode(Base):
         DateTime(timezone=True), nullable=False,
         default=lambda: datetime.now(timezone.utc),
     )
+
+
+class DockerHostStatus(Base):
+    """Whether Docker answered on a host at its agent's last report (agent ≥ 1.9.0).
+
+    "ok" when it did; "unavailable" when there is no Docker to ask (no socket,
+    nothing listening); "error" when Docker is there and the call failed.
+    `detail` says why for the last two. While a host is not "ok" its container
+    rows are what Docker last said, not what is true now, and read as stale.
+    A host that never had Docker gets no row.
+    """
+    __tablename__ = "docker_host_status"
+
+    host_id: Mapped[str] = mapped_column(
+        String(36), ForeignKey("hosts.id", ondelete="CASCADE"), primary_key=True
+    )
+    status: Mapped[str] = mapped_column(String(16), nullable=False)
+    detail: Mapped[str | None] = mapped_column(String(255), nullable=True)
+    # When the status last changed, and the agent's newest report.
+    since: Mapped[datetime] = mapped_column(DateTime(timezone=True), nullable=False)
+    reported_at: Mapped[datetime] = mapped_column(DateTime(timezone=True), nullable=False)

plugins/docker/overview.py

@@ -0,0 +1,150 @@
+# plugins/docker/overview.py
+"""The Docker page's two tables, as plain rows (#5299, #5301): one row per
+swarm service and one per standalone container, with the search, stack, host
+and status filters and the column sort the page offers. Model-free, so it is
+unit-tested without a database; routes.rows() feeds it.
+
+A stack is how the operator thinks of a group of services. Swarm names a
+stack's services "<stack>_<service>", so a service's stack is the part of its
+name before the first "_"; a standalone container's is its compose project.
+Stacks filter the list rather than section it, as host groups do on Hosts.
+"""
+from __future__ import annotations
+
+from datetime import datetime
+
+from .swarm_view import service_trouble
+
+# Status filter values and their labels; NO_STACK picks rows with no stack.
+DOCKER_STATES = {"problem": "Needs attention", "healthy": "Healthy", "stopped": "Stopped"}
+NO_STACK = "__none__"
+# Each sort and whether it starts descending: the busiest first; "up" starts
+# with the shortest uptime, the thing that most recently restarted.
+DOCKER_SORTS = {"status": False, "name": False, "stack": False, "cpu": True, "mem": True,
+                "up": False}
+
+
+def stack_of(service_name: str) -> str | None:
+    """The stack a swarm service was deployed in: "arr" for "arr_radarr";
+    None for a name with no "_" (a service created on its own)."""
+    stack, sep, rest = service_name.partition("_")
+    return stack if sep and stack and rest else None
+
+
+def _seconds_up(started_at, now: datetime) -> float | None:
+    return None if started_at is None else max((now - started_at).total_seconds(), 0.0)
+
+
+def service_row(s: dict, now: datetime) -> dict:
+    """A built swarm service (swarm_view.build_swarm_services) as a table row.
+    CPU and memory add up the running replicas; "up" is the newest replica's
+    start, so a service that just restarted a task reads as just restarted."""
+    reals = [r for r in s["replicas"] if not r["ghost"]]
+    running = [r for r in reals if (r["status"] or "").lower() == "running"]
+    cpu = [r["cpu_pct"] for r in running if r.get("cpu_pct") is not None]
+    mem = [r["mem_bytes"] for r in running if r.get("mem_bytes") is not None]
+    starts = [r["started_at"] for r in running if r.get("started_at")]
+    started = max(starts) if starts else None
+    return {
+        "kind": "service", "name": s["name"], "stack": stack_of(s["name"]),
+        "mode": s["mode"],
+        "image": s.get("image") or next((r["image"] for r in reals if r.get("image")), None),
+        "running": s["running"], "desired": s["desired"], "state": s["state"],
+        "trouble": service_trouble(s),
+        "nodes": sorted({r["host"] for r in reals}, key=str.lower),
+        "ghost_nodes": sorted({r["host"] for r in s["replicas"] if r["ghost"]}, key=str.lower),
+        "cpu_cores": sum(cpu) / 100 if cpu else None,
+        "mem_bytes": sum(mem) if mem else None,
+        "started_at": started, "up_s": _seconds_up(started, now),
+        "unhealthy": any(r.get("health") == "unhealthy" for r in reals),
+        "failures_24h": s.get("failures_24h"), "crash_looping": s.get("crash_looping"),
+        "last_error": s.get("last_error"), "trouble_since": s.get("trouble_since"),
+    }
+
+
+def container_trouble(c, *, stale: bool, exit_failed: bool) -> int | None:
+    """How bad a standalone container is, on service_trouble's scale: 1
+    unhealthy or restarting, 2 died with a failing exit, 3 its report is
+    stale; None when nothing is wrong (a cleanly stopped one included)."""
+    if c.health == "unhealthy" or c.status == "restarting":
+        return 1
+    if c.status != "running" and exit_failed:
+        return 2
+    if stale:
+        return 3
+    return None
+
+
+def container_row(c, *, label: str, host: str, usage: dict, ports: list, stale: bool,
+                  exit_failed: bool, now: datetime) -> dict:
+    """A standalone container as a table row; `usage` is service.usage()."""
+    running = c.status == "running"
+    return {
+        "kind": "container", "name": label, "container_name": c.name, "host": host,
+        "host_id": c.host_id, "stack": c.compose_project or None, "image": c.image,
+        "status": c.status, "health": c.health, "running": running, "ports": ports,
+        "usage": usage, "cpu_cores": usage.get("cores") if running else None,
+        "mem_bytes": usage.get("mem_bytes") if running else None,
+        "started_at": c.started_at if running else None,
+        "up_s": _seconds_up(c.started_at, now) if running else None,
+        "exit_code": c.exit_code, "oom_killed": c.oom_killed, "exit_failed": exit_failed,
+        "restart_count": c.restart_count or 0, "stale": stale, "scraped_at": c.scraped_at,
+        "trouble": container_trouble(c, stale=stale, exit_failed=exit_failed),
+    }
+
+
+def _hosts(row: dict) -> list[str]:
+    return row["nodes"] + row["ghost_nodes"] if row["kind"] == "service" else [row["host"]]
+
+
+def _stopped(row: dict) -> bool:
+    return (row["running"] == 0) if row["kind"] == "service" else not row["running"]
+
+
+def filter_rows(rows: list[dict], *, q: str = "", stack: str = "", host: str = "",
+                state: str = "") -> list[dict]:
+    """Rows matching a search (name, container name, image or stack, any
+    case), a stack (or NO_STACK), a host or node, and a Status filter; an
+    empty value doesn't filter."""
+    q = (q or "").strip().lower()
+    out = []
+    for r in rows:
+        if q and not any(q in (v or "").lower() for v in
+                         (r["name"], r.get("container_name"), r["image"], r["stack"])):
+            continue
+        if (stack == NO_STACK and r["stack"]) or (stack and stack != NO_STACK and r["stack"] != stack):
+            continue
+        if host and host not in _hosts(r):
+            continue
+        if state == "problem" and r["trouble"] is None:
+            continue
+        if state == "healthy" and (r["trouble"] is not None or _stopped(r)):
+            continue
+        if state == "stopped" and not _stopped(r):
+            continue
+        out.append(r)
+    return out
+
+
+def sort_rows(rows: list[dict], sort: str, desc: bool) -> list[dict]:
+    """Rows in a column's order. Ties keep name order; a row with no value for
+    the column goes last whichever way it runs. "status" is problems first,
+    worst first, then running, then stopped."""
+    by_name = sorted(rows, key=lambda r: r["name"].lower())
+    if sort == "name":
+        return list(reversed(by_name)) if desc else by_name
+    if sort == "status":
+        return sorted(by_name, key=lambda r: (4 if r["trouble"] is None else r["trouble"],
+                                              _stopped(r)), reverse=desc)
+    value = {"stack": lambda r: (r["stack"] or "").lower() or None,
+             "cpu": lambda r: r["cpu_cores"], "mem": lambda r: r["mem_bytes"],
+             "up": lambda r: r["up_s"]}[sort]
+    have = [r for r in by_name if value(r) is not None]
+    return sorted(have, key=value, reverse=desc) + [r for r in by_name if value(r) is None]
+
+
+def filter_options(services: list[dict], containers: list[dict]) -> dict:
+    """The stack and host choices the filter bar offers, from every row."""
+    rows = services + containers
+    return {"stacks": sorted({r["stack"] for r in rows if r["stack"]}, key=str.lower),
+            "hosts": sorted({h for r in rows for h in _hosts(r)}, key=str.lower)}

plugins/docker/retention.py

@@ -1,128 +1,95 @@
 # plugins/docker/retention.py
-"""Bound Docker time-series growth: roll up old metrics, prune old rows.
+"""Bound Docker history growth: prune old lifecycle events and log lines.
 
-Published as the "docker.run_retention" capability (see __init__.setup) so the
-core cleanup task can drive it WITHOUT importing the docker models (same
-opportunistic-coupling pattern as docker.persist_host_samples). Runs inside the
-caller's open transaction; never opens or commits its own.
+Published as the "docker.run_retention" and "docker.prune_logs" capabilities
+(see __init__.setup) so the core cleanup task can drive them WITHOUT importing
+the docker models (same opportunistic-coupling pattern as
+docker.persist_host_samples). Each runs inside the caller's open transaction;
+never opens or commits its own.
 
-The scaling concern is docker_metrics: ~2880 rows/container/day at a 30s sample.
-We keep raw samples for a short window, then aggregate everything older into
-hourly averages (docker_metrics_hourly) and delete the raw rows — so multi-day
-history stays cheap to store and query. docker_events is light but unbounded
-without a cutoff, so it gets a (longer) window too.
+Container CPU and memory are not here: they live in the one metric store
+(plugin_metrics) and its rollup and retention (#5752).
 """
 from __future__ import annotations
 
 from datetime import datetime, timedelta
 
 
-def _hour_floor(dt: datetime) -> datetime:
-    """Truncate a datetime down to the start of its hour (drops min/sec/µs)."""
-    return dt.replace(minute=0, second=0, microsecond=0)
-
-
-def _rollup_cutoff(now: datetime, raw_days: int) -> datetime:
-    """Hour-aligned boundary below which raw metrics get rolled up + deleted.
-
-    Aligning to the hour means we only ever roll up *whole* elapsed hours — a
-    bucket is never split across the keep/roll boundary, so re-running can't
-    produce a partial-then-complete duplicate for the same hour.
-    """
-    return _hour_floor(now - timedelta(days=raw_days))
-
-
 async def run_docker_retention(
     session,
     *,
     events_days: int,
-    metrics_raw_days: int,
-    metrics_rollup_days: int,
     now: datetime | None = None,
 ) -> dict:
-    """Roll up + prune Docker time-series. Returns a counts dict for logging.
+    """Prune docker_events older than the events window. Returns a counts
+    dict for logging. The container-log ring is prune_docker_logs, run in its
+    own transaction."""
+    from datetime import timezone
 
-    1. Aggregate docker_metrics older than the (hour-aligned) raw window into
-       docker_metrics_hourly (avg cpu/mem per container per hour), upserting so a
-       re-run is idempotent, then delete those raw rows.
-    2. Prune rolled-up rows older than the rollup window.
-    3. Prune docker_events older than the events window.
+    from sqlalchemy import delete
+
+    from .models import DockerEvent
+
+    if now is None:
+        now = datetime.now(timezone.utc)
+    res = await session.execute(
+        delete(DockerEvent).where(DockerEvent.at < now - timedelta(days=events_days))
+    )
+    return {"events_pruned": res.rowcount or 0}
+
+
+async def prune_docker_logs(
+    session,
+    *,
+    retention_days: int,
+    max_bytes_per_container: int,
+    now: datetime | None = None,
+) -> dict:
+    """The docker_logs ring (m79): drop lines older than the age window, then
+    keep only the newest ~cap bytes per (host, container). Runs inside the
+    caller's open transaction, like run_docker_retention.
+
+    Only containers at or over the cap are ranked (#5449). Ranking every line
+    in the table each hour to find the few containers over it sorted the whole
+    table on slow storage.
     """
     from datetime import timezone
 
-    from sqlalchemy import delete, func, select
-    from sqlalchemy.dialects.postgresql import insert as pg_insert
+    from sqlalchemy import and_, delete, func, select
 
-    from .models import DockerEvent, DockerMetric, DockerMetricHourly
+    from .models import DockerLog
 
     if now is None:
         now = datetime.now(timezone.utc)
 
-    rolled = rolled_rows = events_pruned = rollup_pruned = 0
-
-    # ── 1. Roll up raw metrics older than the raw window into hourly buckets ──
-    raw_cutoff = _rollup_cutoff(now, metrics_raw_days)
-    hour = func.date_trunc("hour", DockerMetric.scraped_at)
-    agg = (
-        select(
-            DockerMetric.host_id,
-            DockerMetric.container_name,
-            hour.label("bucket"),
-            func.avg(DockerMetric.cpu_pct).label("cpu_pct"),
-            func.avg(DockerMetric.mem_pct).label("mem_pct"),
-            func.avg(DockerMetric.mem_usage_bytes).label("mem_usage_bytes"),
-            func.count().label("sample_count"),
-        )
-        .where(DockerMetric.scraped_at < raw_cutoff)
-        .group_by(DockerMetric.host_id, DockerMetric.container_name, hour)
-    )
-    for r in (await session.execute(agg)).all():
-        stmt = (
-            pg_insert(DockerMetricHourly)
-            .values(
-                host_id=r.host_id,
-                container_name=r.container_name,
-                bucket=r.bucket,
-                cpu_pct=float(r.cpu_pct or 0.0),
-                mem_pct=float(r.mem_pct or 0.0),
-                mem_usage_bytes=int(r.mem_usage_bytes or 0),
-                sample_count=int(r.sample_count or 0),
-            )
-            .on_conflict_do_update(
-                constraint="uq_docker_metrics_hourly_bucket",
-                set_={
-                    "cpu_pct": float(r.cpu_pct or 0.0),
-                    "mem_pct": float(r.mem_pct or 0.0),
-                    "mem_usage_bytes": int(r.mem_usage_bytes or 0),
-                    "sample_count": int(r.sample_count or 0),
-                },
-            )
-        )
-        await session.execute(stmt)
-        rolled += 1
-        rolled_rows += int(r.sample_count or 0)
-    if rolled:
-        await session.execute(
-            delete(DockerMetric).where(DockerMetric.scraped_at < raw_cutoff)
-        )
-
-    # ── 2. Prune rolled-up rows beyond the rollup window ──
-    rollup_cutoff = now - timedelta(days=metrics_rollup_days)
     res = await session.execute(
-        delete(DockerMetricHourly).where(DockerMetricHourly.bucket < rollup_cutoff)
+        delete(DockerLog).where(DockerLog.ts < now - timedelta(days=retention_days))
     )
-    rollup_pruned = res.rowcount or 0
+    age_pruned = res.rowcount or 0
 
-    # ── 3. Prune lifecycle events beyond the events window ──
-    events_cutoff = now - timedelta(days=events_days)
+    size = func.length(DockerLog.line)
+    over = (
+        select(DockerLog.host_id, DockerLog.container_name)
+        .group_by(DockerLog.host_id, DockerLog.container_name)
+        .having(func.sum(size) >= max_bytes_per_container)
+        .subquery()
+    )
+    # Per container, sum line bytes newest-first; delete a row once its
+    # strictly-newer siblings already fill the cap. The EXCLUSIVE prefix
+    # (running total minus this row) means the newest row always survives, so
+    # a single line larger than the cap is never wiped out.
+    running = func.sum(size).over(
+        partition_by=[DockerLog.host_id, DockerLog.container_name],
+        order_by=[DockerLog.ts.desc(), DockerLog.id.desc()],
+    )
+    ranked = (
+        select(DockerLog.id, (running - size).label("prefix_excl"))
+        .join(over, and_(DockerLog.host_id == over.c.host_id,
+                         DockerLog.container_name == over.c.container_name))
+        .subquery()
+    )
     res = await session.execute(
-        delete(DockerEvent).where(DockerEvent.at < events_cutoff)
+        delete(DockerLog).where(DockerLog.id.in_(
+            select(ranked.c.id).where(ranked.c.prefix_excl >= max_bytes_per_container)))
     )
-    events_pruned = res.rowcount or 0
-
-    return {
-        "buckets_rolled": rolled,
-        "raw_rows_rolled": rolled_rows,
-        "rollup_pruned": rollup_pruned,
-        "events_pruned": events_pruned,
-    }
+    return {"logs_age_pruned": age_pruned, "logs_size_pruned": res.rowcount or 0}

plugins/docker/service.py

@@ -0,0 +1,750 @@
+"""Docker queries shared by the plugin's pages and its MCP tools.
+
+Each function takes an open session and returns rows or plain data, with no
+Quart request context, so the routes and the /mcp tools (which run outside
+Quart's pipeline) call the same code.
+"""
+from __future__ import annotations
+
+import re
+from datetime import datetime, timedelta, timezone
+
+from sqlalchemy import func, select
+
+from steward.core import metric_series
+from steward.models.hosts import Host
+from steward.models.metrics import PluginMetric
+
+from .dedup import dedup_by_container_id
+from .models import (
+    DockerContainer, DockerEvent, DockerLog, DockerSwarmService,
+)
+from .swarm_view import replaced_tasks
+
+
+CRASH_LOOP_DEATHS = 3  # default; Settings → Thresholds sets it
+
+
+def crash_loop_deaths(config) -> int:
+    """Deaths within an hour that make a container or a swarm service a crash
+    loop, from Settings → Thresholds (app.config["THRESHOLDS"])."""
+    th = (config.get("THRESHOLDS") or {}).get("crash_loop") or {}
+    return int(th.get("warn") or CRASH_LOOP_DEATHS)
+
+
+# Exit codes a deliberate stop leaves: SIGTERM (143), and SIGKILL (137) when
+# the process outlives the stop timeout. Without an OOM kill they are someone
+# stopping the container, not the container failing.
+STOP_EXITS = (137, 143)
+
+
+def exit_is_failure(exit_code, oom_killed: bool = False) -> bool:
+    """Whether a container's exit is a failure: an OOM kill, or a non-zero
+    exit other than a stop's."""
+    return bool(oom_killed) or exit_code not in (None, 0, *STOP_EXITS)
+
+
+def _event_exit(detail) -> int | None:
+    """The exit code in a die event's detail ("exit 1"), or None."""
+    text = (detail or "").strip()
+    if text.startswith("exit "):
+        try:
+            return int(text[5:])
+        except ValueError:
+            return None
+    return None
+
+
+def event_is_failure(e) -> bool:
+    """Whether a die/oom event is a failure. An oom always is; a die is unless
+    its exit is a stop's (or clean)."""
+    if e.event == "oom":
+        return True
+    return e.event == "die" and exit_is_failure(_event_exit(e.detail))
+
+
+def failures_24h(events, swarm_services, host_name: dict) -> list[dict]:
+    """What failed in the last 24h, newest first, for the dashboard count and
+    the Docker page's list of it: one row per plain container that died or was
+    OOM-killed, and one per swarm service with failed tasks.
+
+    `events` are die/oom events of the last 24h. A swarm task's container dies
+    whenever Swarm replaces it, so a service with Swarm's task records counts
+    its failed tasks instead of its containers' deaths; a service without them
+    (an agent older than 1.8.0) falls back to its tasks' deaths.
+
+    Each row: {kind: "container"|"service", name, host_id, host, count, at,
+    detail, task_id}; task_id is the failed swarm task, so its logs can be
+    found (None for a container, or a service without task records).
+    """
+    by_prefix = {s["name"] + ".": s for s in swarm_services}
+    rows: dict[tuple, dict] = {}
+    # Each manager reports every swarm task, so one task's death can arrive
+    # once per manager: a service counts each dead task container once.
+    task_deaths: dict[tuple, set] = {}
+    for e in sorted(events, key=lambda e: e.at, reverse=True):
+        if not event_is_failure(e):
+            continue
+        svc = next((s for p, s in by_prefix.items() if e.container_name.startswith(p)), None)
+        if svc is not None and svc["tasks_known"]:
+            continue
+        kind, name = ("service", svc["name"]) if svc else ("container", e.container_name)
+        key = (kind, name) if svc else (kind, e.host_id, name)
+        if svc is not None:
+            seen = task_deaths.setdefault(key, set())
+            if e.container_name in seen:
+                continue
+            seen.add(e.container_name)
+        row = rows.get(key)
+        if row is None:
+            rows[key] = {"kind": kind, "name": name, "host_id": e.host_id,
+                         "host": host_name.get(e.host_id, e.host_id), "count": 1, "at": e.at,
+                         "detail": "OOM-killed" if e.event == "oom" else e.detail,
+                         "task_id": (task_of(svc["name"], e.container_name) or (None, None))[1]
+                         if svc else None}
+        else:
+            row["count"] += 1
+    from .swarm_view import task_time
+    for s in swarm_services:
+        if s["tasks_known"] and s["failures_24h"]:
+            err = s["last_error"] or {}
+            rows[("service", s["name"])] = {
+                "kind": "service", "name": s["name"], "host_id": None, "host": err.get("node"),
+                "count": s["failures_24h"], "at": task_time(err) if err else None,
+                "detail": err.get("error") or err.get("message") or err.get("state"),
+                "task_id": err.get("task_id")}
+    epoch = datetime.min.replace(tzinfo=timezone.utc)
+    return sorted(rows.values(), key=lambda r: r["at"] or epoch, reverse=True)
+
+
+def task_label(c, node_hostname: dict | None = None) -> str:
+    """How a container reads to a person. A swarm task is its service and
+    slot: "codex_db · 1" for a replica, "agent · node-3" for a global
+    service's task on a node. Anything else as display_name reads it."""
+    from .swarm_view import _slot_key
+    svc = getattr(c, "service_name", None)
+    if not svc:
+        return display_name(c)
+    slot = _slot_key(c)
+    if slot.isdigit():
+        return f"{svc} · {slot}"
+    host = (node_hostname or {}).get(slot) or (node_hostname or {}).get(getattr(c, "node_id", None))
+    return f"{svc} · {host}" if host else svc
+
+
+_CI_JOB = re.compile(
+    r"^(?:GITEA|FORGEJO)-ACTIONS-TASK-\d+-WORKFLOW-(?P<workflow>.+?)-JOB-(?P<job>.+?)(?:-[0-9a-f]{16,})?$")
+# docker buildx's container driver names its node "buildx_buildkit_<builder><n>".
+_BUILDKIT = re.compile(r"^buildx_buildkit_(?P<builder>.+?)\d*$")
+
+
+def display_name(c) -> str:
+    """A container's name as a list can show it in one short line (#5253): a
+    swarm task by its service (the host says where it runs), a Gitea or
+    Forgejo Actions job container as "CI: <workflow> · <job>", a buildx
+    builder as "BuildKit builder · <name>", anything else by its own name."""
+    svc = getattr(c, "service_name", None)
+    if svc:
+        return svc
+    m = _CI_JOB.match(c.name or "")
+    if m:
+        return f"CI: {m['workflow']} · {m['job']}"
+    m = _BUILDKIT.match(c.name or "")
+    if m:
+        return f"BuildKit builder · {m['builder']}"
+    return c.name
+
+
+def usage(c, cpus: int | None, mem_total: int | None, cpu_pct: float | None = None, *,
+          thresholds: dict | None = None, norms: dict | None = None) -> dict:
+    """A container's CPU and memory in terms a reader can weigh.
+
+    Docker's CPU% is per core (100% = one core busy), so it is shown as cores
+    and graded as a share of the host's CPUs when the host's core count is
+    known. Docker's memory limit is the host's RAM when none is set, so memory
+    is shown in bytes, against a limit only when one is actually set; mem_pct
+    is then a share of that limit, otherwise of the host's memory. `cpu_pct`
+    replaces the container's last sample (an average, say) when given.
+
+    With `thresholds`, cpu_grade and mem_grade are the grades to show: past
+    the cutoff but within the container's own normal for the hour (`norms`,
+    {metric: Norm}, container_norms) is "ok", and cpu_usual / mem_usual say
+    what is usual (#5742). Without, they are None and the page grades.
+    """
+    cpu = getattr(c, "cpu_pct", None) if cpu_pct is None else cpu_pct
+    limit = getattr(c, "mem_limit_bytes", None)
+    limited = bool(limit and mem_total and limit < mem_total * 0.95)
+    out = {
+        "cores": None if cpu is None else cpu / 100.0,
+        "cpu_share": None if cpu is None or not cpus else cpu / cpus,
+        "cpus": cpus,
+        "mem_bytes": getattr(c, "mem_usage_bytes", None),
+        "mem_limit": limit if limited else None,
+        "mem_pct": getattr(c, "mem_pct", None),
+        "mem_of": "its limit" if limited else ("the host's memory" if mem_total else None),
+        "cpu_grade": None, "cpu_usual": None, "mem_grade": None, "mem_usual": None,
+    }
+    if thresholds is not None:
+        _grade(out, cpu, thresholds, norms or {})
+    return out
+
+
+# Container figures graded against their own normal (#5742): the metric that
+# learns it and its floor's setting (norms.floor.*).
+NORMED = {"cpu": ("cpu_pct", "container_cpu_pct"), "mem": ("mem_pct", "container_mem_pct")}
+
+
+def _grade(u: dict, cpu: float | None, thresholds: dict, norms: dict) -> None:
+    """Fill usage()'s grades. CPU is graded as a share of the host but learned
+    per core (Docker's own figure), so its normal is read in cores."""
+    from steward.core.anomalies import departs, graded
+    from steward.core.formatting import status_class
+    from steward.core.settings import floor_of
+    grade = status_class(u["cpu_share"], "cpu", thresholds)
+    norm = norms.get("cpu_pct")
+    if grade in ("warn", "crit") and norm is not None and cpu is not None \
+            and not departs(cpu, norm, floor_of(thresholds, "container_cpu_pct")):
+        grade = "ok"
+        u["cpu_usual"] = f"usual {norm.usual_hold / 100:.2f} cores at this hour"
+    u["cpu_grade"] = grade
+    if u["mem_of"]:
+        u["mem_grade"], u["mem_usual"] = graded(
+            u["mem_pct"], "mem", thresholds, norms.get("mem_pct"),
+            floor_of(thresholds, "container_mem_pct"), "%")
+    else:
+        u["mem_grade"] = "dim"
+
+
+async def container_norms(db, keys, now: datetime | None = None) -> dict:
+    """{(host_id, container_name): {metric: Norm}}: each container's CPU and
+    memory normals for this hour. One without enough history is absent."""
+    from steward.core.norms import norms_at
+    resources = await container_resources(db, keys)
+    got = await norms_at(db, [("docker", r, metric) for r in resources
+                              for metric, _f in NORMED.values()],
+                         now or datetime.now(timezone.utc))
+    out: dict = {}
+    for (_s, resource, metric), norm in got.items():
+        out.setdefault(resources[resource], {})[metric] = norm
+    return out
+
+
+async def host_capacity(db, host_ids) -> dict[str, tuple[int | None, int | None]]:
+    """{host_id: (CPU cores, memory bytes)} from the host agent's latest
+    readings: cores counted from its per-core CPU series, memory from
+    available ÷ (1 − used%). (None, None) for a host without an agent."""
+    from sqlalchemy import or_
+    from steward.models.metrics import PluginMetricLatest
+    hosts = await host_map(db, set(host_ids))
+    if not hosts:
+        return {}
+    names = {h.name: hid for hid, h in hosts.items()}
+    rows = (await db.execute(
+        select(PluginMetricLatest.resource_name, PluginMetricLatest.metric_name,
+               PluginMetricLatest.value)
+        .where(PluginMetricLatest.source_module == "host_agent",
+               or_(PluginMetricLatest.resource_name.in_(list(names)),
+                   *(PluginMetricLatest.resource_name.like(escape_like(n) + ":core%", escape="\\")
+                     for n in names)))
+    )).all()
+    cores: dict[str, int] = {}
+    mem: dict[str, dict] = {}
+    for res, metric, value in rows:
+        if res in names:
+            mem.setdefault(res, {})[metric] = value
+        elif metric == "cpu_pct":
+            host = res.rsplit(":core", 1)[0]
+            cores[host] = cores.get(host, 0) + 1
+    out = {}
+    for name, hid in names.items():
+        m = mem.get(name, {})
+        avail, used = m.get("mem_available_bytes"), m.get("mem_used_pct")
+        total = int(avail / (1 - used / 100)) if avail and used is not None and used < 99.5 else None
+        out[hid] = (cores.get(name), total)
+    return out
+
+
+async def host_map(db, host_ids: set[str]) -> dict[str, Host]:
+    if not host_ids:
+        return {}
+    return {
+        h.id: h for h in (await db.execute(
+            select(Host).where(Host.id.in_(host_ids)))).scalars().all()
+    }
+
+
+def escape_like(text: str) -> str:
+    """Make `text` match literally inside a LIKE/ILIKE pattern (escape char \\)."""
+    return text.replace("\\", "\\\\").replace("%", "\\%").replace("_", "\\_")
+
+
+async def list_containers(db, host_id: str | None = None) -> list[DockerContainer]:
+    """Containers, running first then by name; one host's when host_id is given."""
+    stmt = select(DockerContainer)
+    if host_id is not None:
+        stmt = stmt.where(DockerContainer.host_id == host_id)
+    stmt = stmt.order_by((DockerContainer.status == "running").desc(), DockerContainer.name)
+    return list((await db.execute(stmt)).scalars())
+
+
+async def container_events(db, host_id: str, name: str, limit: int = 50) -> list[DockerEvent]:
+    """A container's lifecycle events (start/stop/die/oom/health), newest first."""
+    return list((await db.execute(
+        select(DockerEvent)
+        .where(DockerEvent.host_id == host_id)
+        .where(DockerEvent.container_name == name)
+        .order_by(DockerEvent.at.desc())
+        .limit(limit)
+    )).scalars())
+
+
+async def query_logs(db, host_id: str, name: str, stream: str, query: str,
+                     limit: int) -> list:
+    """The recent retained lines for one container, newest-first, optionally
+    filtered by stream and a case-insensitive substring.
+
+    Newest-first is deliberate: the viewer replaces the list on each poll, which
+    resets scroll to the top — so the latest lines stay visible without any
+    scroll handling, and older lines are a scroll away.
+    """
+    stmt = (select(DockerLog.ts, DockerLog.stream, DockerLog.line)
+            .where(DockerLog.host_id == host_id)
+            .where(DockerLog.container_name == name))
+    if stream in ("stdout", "stderr"):
+        stmt = stmt.where(DockerLog.stream == stream)
+    if query:
+        stmt = stmt.where(DockerLog.line.ilike(f"%{escape_like(query)}%", escape="\\"))
+    stmt = stmt.order_by(DockerLog.ts.desc(), DockerLog.id.desc()).limit(limit)
+    return (await db.execute(stmt)).all()
+
+
+CONTAINER_SERIES = ("cpu_pct", "mem_pct", "mem_usage_bytes")
+
+
+# A container's series live in the one metric store (#5752): source "docker",
+# resource "<host name>/<container name>", recorded at ingest.
+def container_resource(host_name: str, name: str) -> str:
+    return f"{host_name}/{name}"
+
+
+async def container_resources(db, keys) -> dict[str, tuple[str, str]]:
+    """{resource: (host_id, container_name)} for (host_id, container_name)
+    keys whose host still exists."""
+    keys = set(keys)
+    hosts = await host_map(db, {h for h, _ in keys})
+    return {container_resource(hosts[h].name, n): (h, n) for h, n in keys if h in hosts}
+
+
+async def container_points(db, host_id: str, name: str, metric: str, since,
+                           target: int = 120) -> list[tuple]:
+    """(time, average) points of one container metric since `since`, oldest
+    first: raw samples bucket-averaged, and the hourly rollup for the part of
+    the range past the raw window (metric_series.points)."""
+    if metric not in CONTAINER_SERIES:
+        raise ValueError(f"metric must be one of {', '.join(CONTAINER_SERIES)}")
+    host = await db.get(Host, host_id)
+    if host is None:
+        return []
+    return await metric_series.points(db, "docker", container_resource(host.name, name),
+                                      metric, since, target=target)
+
+
+async def _samples(db, resources, metrics, since: datetime, width: int | None = None):
+    """Rows of (resource, time, metric, value) for the given containers since
+    `since`: each sample, or with `width` (seconds) each bucket's average."""
+    at = PluginMetric.recorded_at
+    if width:
+        at = func.date_bin(func.make_interval(0, 0, 0, 0, 0, 0, int(width)),
+                           PluginMetric.recorded_at, func.to_timestamp(0))
+    value = func.avg(PluginMetric.value) if width else PluginMetric.value
+    q = (select(PluginMetric.resource_name, at.label("at"), PluginMetric.metric_name, value)
+         .where(PluginMetric.source_module == "docker",
+                PluginMetric.resource_name.in_(list(resources)),
+                PluginMetric.metric_name.in_(list(metrics)),
+                PluginMetric.recorded_at >= since))
+    if width:
+        q = q.group_by(PluginMetric.resource_name, at, PluginMetric.metric_name)
+    return (await db.execute(q.order_by(at))).all()
+
+
+async def average_cpu(db, keys, since: datetime) -> dict[tuple[str, str], float]:
+    """{(host_id, container_name): its average cpu_pct since `since`}."""
+    resources = await container_resources(db, keys)
+    if not resources:
+        return {}
+    rows = (await db.execute(
+        select(PluginMetric.resource_name, func.avg(PluginMetric.value))
+        .where(PluginMetric.source_module == "docker",
+               PluginMetric.resource_name.in_(list(resources)),
+               PluginMetric.metric_name == "cpu_pct",
+               PluginMetric.recorded_at >= since)
+        .group_by(PluginMetric.resource_name))).all()
+    return {resources[r]: float(v) for r, v in rows if v is not None}
+
+
+# Busiest Containers bars (#5214): the recent past in BARS equal slices.
+BARS = 24
+
+
+def bar_slots(now: datetime, hours: int, bars: int = BARS) -> tuple[list[datetime], int]:
+    """The start of each of the last `bars` buckets covering `hours`, oldest
+    first, aligned as date_bin aligns them (to the epoch), and their width in
+    seconds. The newest bucket is the one `now` falls in."""
+    width = max(60, hours * 3600 // bars)
+    last = int(now.timestamp()) // width * width
+    return [datetime.fromtimestamp(last - (bars - 1 - i) * width, timezone.utc)
+            for i in range(bars)], width
+
+
+async def recent_usage(db, keys, since: datetime, width: int) -> dict:
+    """{(host_id, container_name): {bucket start: (cpu_pct, mem_usage_bytes)}}
+    averaged per bucket since `since`, for the given containers, in one query."""
+    resources = await container_resources(db, keys)
+    if not resources:
+        return {}
+    out: dict = {}
+    for resource, at, metric, value in await _samples(
+            db, resources, ("cpu_pct", "mem_usage_bytes"), since, width):
+        cpu, mem = out.setdefault(resources[resource], {}).get(at, (None, None))
+        v = None if value is None else float(value)
+        out[resources[resource]][at] = (v, mem) if metric == "cpu_pct" else (cpu, v)
+    return out
+
+
+def scaled_bars(values: list, floor: float) -> dict:
+    """Bar heights in % of a roof: the row's own peak, but never below `floor`,
+    so a container idling at 0.05 cores draws flat and its 0.05 → 0.10 wobble
+    doesn't look like a 1.5 → 3 core spike. None (no samples) stays a gap."""
+    present = [v for v in values if v is not None]
+    peak = max(present) if present else None
+    roof = max(peak or 0.0, floor)
+    return {"bars": [None if v is None else round(100.0 * v / roof, 1) for v in values],
+            "peak": peak, "roof": roof}
+
+
+async def search_logs(db, *, since, host_id: str | None = None, name: str | None = None,
+                      containers: list[tuple[str, str]] | None = None,
+                      text: str = "", stream: str = "", before=None, limit: int = 50) -> list:
+    """Retained log lines matching a filter, newest first (limit + 1 fetched by callers
+    that need to know whether more exist). `containers` keeps the lines of those
+    (host_id, container_name) pairs: a swarm service's tasks (#5468)."""
+    from sqlalchemy import tuple_
+    stmt = (select(DockerLog.id, DockerLog.host_id, DockerLog.container_name,
+                   DockerLog.ts, DockerLog.stream, DockerLog.line)
+            .where(DockerLog.ts >= since))
+    if host_id:
+        stmt = stmt.where(DockerLog.host_id == host_id)
+    if name:
+        stmt = stmt.where(DockerLog.container_name == name)
+    if containers is not None:
+        if not containers:
+            return []
+        stmt = stmt.where(tuple_(DockerLog.host_id, DockerLog.container_name).in_(containers))
+    if stream in ("stdout", "stderr"):
+        stmt = stmt.where(DockerLog.stream == stream)
+    if text:
+        stmt = stmt.where(DockerLog.line.ilike(f"%{escape_like(text)}%", escape="\\"))
+    if before is not None:
+        stmt = stmt.where(DockerLog.ts < before)
+    stmt = stmt.order_by(DockerLog.ts.desc(), DockerLog.id.desc()).limit(limit)
+    return list((await db.execute(stmt)).all())
+
+
+async def log_context(db, host_id: str, name: str, ts, n: int) -> tuple[list, list]:
+    """Up to n lines either side of `ts` in one container's log, each oldest first."""
+    cols = (DockerLog.id, DockerLog.ts, DockerLog.stream, DockerLog.line)
+    base = select(*cols).where(DockerLog.host_id == host_id, DockerLog.container_name == name)
+    before = (await db.execute(
+        base.where(DockerLog.ts < ts).order_by(DockerLog.ts.desc(), DockerLog.id.desc()).limit(n)
+    )).all()
+    after = (await db.execute(
+        base.where(DockerLog.ts > ts).order_by(DockerLog.ts, DockerLog.id).limit(n)
+    )).all()
+    return list(reversed(before)), list(after)
+
+
+async def recent_failures_by_container(db, since) -> dict[tuple[str, str], list]:
+    """{(host_id, container_name): [die/oom events since `since`, newest first]},
+    leaving out a die that was a stop (event_is_failure)."""
+    rows = (await db.execute(
+        select(DockerEvent)
+        .where(DockerEvent.event.in_(("die", "oom")), DockerEvent.at >= since)
+        .order_by(DockerEvent.at.desc())
+    )).scalars()
+    out: dict[tuple[str, str], list] = {}
+    for e in rows:
+        if event_is_failure(e):
+            out.setdefault((e.host_id, e.container_name), []).append(e)
+    return out
+
+
+async def failure_events(db, since) -> list:
+    """Die and OOM events since `since`, newest first; failures_24h() decides
+    which of them are failures."""
+    return list((await db.execute(
+        select(DockerEvent)
+        .where(DockerEvent.event.in_(("die", "oom")), DockerEvent.at >= since)
+        .order_by(DockerEvent.at.desc())
+    )).scalars())
+
+
+async def retained_names(db, want: str, *, exact: bool, host_id: str | None = None,
+                         limit: int = 20) -> list[tuple[str, str, datetime]]:
+    """(host_id, container_name, last seen) of containers Steward still holds
+    logs or events for, newest first — including ones Docker has removed.
+
+    exact: the name itself, from both logs and events. Otherwise names that
+    contain `want` (case-insensitive), from events only: a substring match
+    over every retained log line's name would scan the log table, and a
+    container Steward watched die has a die or stop event.
+    """
+    seen: dict[tuple[str, str], datetime] = {}
+    sources = ((DockerLog, DockerLog.ts), (DockerEvent, DockerEvent.at)) if exact \
+        else ((DockerEvent, DockerEvent.at),)
+    for model, at in sources:
+        match = (model.container_name == want if exact
+                 else model.container_name.ilike(f"%{escape_like(want)}%", escape="\\"))
+        stmt = (select(model.host_id, model.container_name, func.max(at))
+                .where(match).group_by(model.host_id, model.container_name))
+        if host_id:
+            stmt = stmt.where(model.host_id == host_id)
+        for hid, name, last in (await db.execute(stmt)).all():
+            key = (hid, name)
+            if key not in seen or last > seen[key]:
+                seen[key] = last
+    rows = sorted(((h, n, t) for (h, n), t in seen.items()), key=lambda r: r[2], reverse=True)
+    return rows[:limit]
+
+
+# A swarm task's container is named "<service>.<slot>.<task id>": slot is the
+# replica number, or the node id for a global service; task ids are 25 chars.
+_TASK_SUFFIX = re.compile(r"^(?P<slot>[A-Za-z0-9]+)\.(?P<task>[a-z0-9]{25})$")
+
+
+def task_of(service: str, container_name: str) -> tuple[str, str] | None:
+    """(slot, task id) when `container_name` is one of `service`'s task
+    containers, else None — so "arr" does not claim "arr_radarr.1.…"."""
+    if not container_name.startswith(service + "."):
+        return None
+    m = _TASK_SUFFIX.match(container_name[len(service) + 1:])
+    return (m["slot"], m["task"]) if m else None
+
+
+async def service_containers(db, service: str) -> list[tuple[str, str]]:
+    """(host_id, container_name) of every task container of a swarm service
+    that Steward still holds anything for: the live ones, and the ones Docker
+    removed that left events (each task container has a start event). The
+    page's logs and history read across all of them (#5302)."""
+    pattern = escape_like(service + ".") + "%"
+    found: set[tuple[str, str]] = set()
+    for model, col in ((DockerContainer, DockerContainer.name),
+                       (DockerEvent, DockerEvent.container_name)):
+        rows = (await db.execute(
+            select(model.host_id, col).where(col.like(pattern, escape="\\")).distinct())).all()
+        found.update((h, n) for h, n in rows if task_of(service, n))
+    return sorted(found)
+
+
+async def query_service_logs(db, containers, stream: str, query: str, limit: int,
+                             task: str | None = None) -> list:
+    """The newest retained lines across a service's task containers, newest
+    first, each with its container name so the page can say which task wrote
+    it. `task` keeps one task's lines (a failure's "Logs →")."""
+    from sqlalchemy import tuple_
+    pairs = [(h, n) for h, n in containers if task is None or n.endswith("." + task)]
+    if not pairs:
+        return []
+    stmt = (select(DockerLog.ts, DockerLog.stream, DockerLog.line, DockerLog.container_name)
+            .where(tuple_(DockerLog.host_id, DockerLog.container_name).in_(pairs)))
+    if stream in ("stdout", "stderr"):
+        stmt = stmt.where(DockerLog.stream == stream)
+    if query:
+        stmt = stmt.where(DockerLog.line.ilike(f"%{escape_like(query)}%", escape="\\"))
+    stmt = stmt.order_by(DockerLog.ts.desc(), DockerLog.id.desc()).limit(limit)
+    return (await db.execute(stmt)).all()
+
+
+def sum_series(samples, since: datetime, until: datetime, points: int = 60) -> list:
+    """[(bucket start, cpu %, memory bytes)] summed across containers: each
+    container's samples are averaged within a bucket, then the containers are
+    added, so two replicas read as their combined load. `samples` are
+    (container_name, at, cpu_pct, mem_bytes); buckets with no sample are left
+    out rather than read as zero."""
+    width = max((until - since).total_seconds() / points, 60.0)
+    acc: dict[int, dict[str, list]] = {}
+    for name, at, cpu, mem in samples:
+        i = int((at - since).total_seconds() // width)
+        if i < 0:
+            continue
+        per = acc.setdefault(i, {}).setdefault(name, [0.0, 0, 0])
+        per[0] += cpu or 0.0
+        per[1] += mem or 0
+        per[2] += 1
+    out = []
+    for i in sorted(acc):
+        cpu = sum(c / n for c, _, n in acc[i].values())
+        mem = sum(m / n for _, m, n in acc[i].values())
+        out.append((since + timedelta(seconds=i * width), cpu, int(mem)))
+    return out
+
+
+async def service_history(db, containers, since: datetime) -> list:
+    """sum_series over a service's task containers' samples since `since`."""
+    resources = await container_resources(db, containers)
+    if not resources:
+        return []
+    joined: dict[tuple[str, datetime], list] = {}
+    for resource, at, metric, value in await _samples(
+            db, resources, ("cpu_pct", "mem_usage_bytes"), since):
+        pair = joined.setdefault((resource, at), [None, None])
+        pair[0 if metric == "cpu_pct" else 1] = value
+    rows = [(resource, at, cpu, mem) for (resource, at), (cpu, mem) in
+            sorted(joined.items(), key=lambda kv: kv[0][1])]
+    return sum_series(rows, since, datetime.now(timezone.utc))
+
+
+async def unhealthy_since(db, keys) -> dict[tuple[str, str], datetime]:
+    """{(host_id, container_name): when it last turned unhealthy} for the given
+    containers, from their health_change events; absent when none is retained."""
+    keys = set(keys)
+    if not keys:
+        return {}
+    rows = (await db.execute(
+        select(DockerEvent)
+        .where(DockerEvent.event == "health_change",
+               DockerEvent.detail.like("%→unhealthy"),
+               DockerEvent.container_name.in_({name for _, name in keys}))
+        .order_by(DockerEvent.at.desc())
+    )).scalars()
+    out: dict[tuple[str, str], datetime] = {}
+    for e in rows:
+        key = (e.host_id, e.container_name)
+        if key in keys and key not in out:
+            out[key] = e.at
+    return out
+
+
+PRUNE_TARGETS = ("containers", "images", "system")
+
+
+def prune_extra_vars(prune_target: str) -> dict:
+    """Extra-vars for the bundled docker_prune playbook. "Prune unused images"
+    means ALL unused (docker image prune -a), not just dangling; the full system
+    prune stays conservative (-f, no -a) per the operator's choice (m78)."""
+    extra_vars: dict = {"prune_target": prune_target}
+    if prune_target == "images":
+        extra_vars["prune_all_images"] = True
+    return extra_vars
+
+
+# ── what is wrong with a container ───────────────────────────────────────────
+# Shared by the Docker page's "Needs attention" and the MCP tools, so both
+# rank the same containers the same way.
+
+PROBLEMS = ("restarting", "crash_looping", "unhealthy", "oom_killed", "exited_nonzero")
+
+def container_problems(c, failures_1h: int, failed_24h: bool,
+                       crash_loop_deaths: int = CRASH_LOOP_DEATHS) -> list[str]:
+    """What is wrong with one container, most severe first; [] when nothing is."""
+    status = (c.status or "").lower()
+    out = []
+    if status == "restarting":
+        out.append("restarting")
+    if failures_1h >= crash_loop_deaths:
+        out.append("crash_looping")
+    if status == "running" and (c.health or "") == "unhealthy":
+        out.append("unhealthy")
+    if c.oom_killed:
+        out.append("oom_killed")
+    if status in ("exited", "dead") and exit_is_failure(c.exit_code) and failed_24h:
+        out.append("exited_nonzero")
+    return out
+
+
+def problem_since(problems: list[str], deaths: list, hour_ago: datetime,
+                  unhealthy_at: datetime | None = None) -> datetime | None:
+    """When a container's worst problem began; None when the retained events
+    cannot say. `deaths` are its die/oom events of the last 24h, newest first.
+
+    A restart or crash loop dates from its first death in the last hour, an
+    OOM kill or error exit from the death that ended it, and a failing
+    healthcheck from when its health last turned unhealthy.
+    """
+    if not problems:
+        return None
+    worst = problems[0]
+    if worst in ("restarting", "crash_looping"):
+        in_hour = [e.at for e in deaths if e.at >= hour_ago]
+        return min(in_hour) if in_hour else (deaths[0].at if deaths else None)
+    if worst == "unhealthy":
+        return unhealthy_at
+    if worst == "oom_killed":
+        ooms = [e.at for e in deaths if e.event == "oom"]
+        return ooms[0] if ooms else (deaths[0].at if deaths else None)
+    return deaths[0].at if deaths else None
+
+
+PROBLEM_SEVERITY = {"restarting": "critical", "crash_looping": "critical", "unhealthy": "critical",
+             "oom_killed": "warning", "exited_nonzero": "warning"}
+PROBLEM_SAYS = {"restarting": "is restarting", "crash_looping": "keeps dying",
+         "unhealthy": "fails its healthcheck", "oom_killed": "was OOM-killed",
+         "exited_nonzero": "exited with an error"}
+
+
+
+async def ranked_containers(db, now: datetime, crash_loop_deaths: int) -> list[dict]:
+    """Every container (deduplicated) with what is wrong with it, most severe
+    first: {container, problems, since, replaced}. A swarm task Swarm has
+    replaced has no problems; its failure belongs to its service."""
+    containers = dedup_by_container_id(await list_containers(db))
+    services = list((await db.execute(select(DockerSwarmService))).scalars())
+    replaced = replaced_tasks(containers, services)
+    failures = await recent_failures_by_container(db, now - timedelta(hours=24))
+    hour_ago = now - timedelta(hours=1)
+    sick = await unhealthy_since(
+        db, [(c.host_id, c.name) for c in containers if (c.health or "") == "unhealthy"])
+    out = []
+    for c in containers:
+        key = (c.host_id, c.name)
+        gone = key in replaced
+        events = failures.get(key, [])
+        recent = sum(1 for e in events if e.at >= hour_ago)
+        problems = [] if gone else container_problems(c, recent, bool(events), crash_loop_deaths)
+        out.append({"container": c, "problems": problems, "replaced": gone,
+                    "since": problem_since(problems, events, hour_ago, sick.get(key))})
+    return out
+
+
+def needs_attention(ranked: list[dict]) -> list[dict]:
+    """The containers with something wrong, worst problem first, then by host
+    and name."""
+    return sorted((r for r in ranked if r["problems"]),
+                  key=lambda r: (PROBLEMS.index(r["problems"][0]), r["container"].host_id, r["container"].name))
+
+
+def stale_after_seconds(config) -> int:
+    """How old a container's last report may be before the page calls it
+    stale: the host agent's own staleness window, since the agent reports it."""
+    return int(((config.get("PLUGINS") or {}).get("host_agent") or {}).get("stale_after_seconds", 180))
+
+
+async def disk_hosts(db, since: datetime, *, points: int = 60) -> list[dict]:
+    """Each Docker host's disk as `docker system df` last said, most space
+    first, with its recorded total since `since` (#5304): the points, oldest
+    first, and the growth over them (disk_view.growth; None until there are
+    two). Hosts are named by their Steward host."""
+    from steward.core import metric_series
+    from .disk_view import DISK_PARTS, growth
+    from .models import DockerDiskUsage
+    summaries = list((await db.execute(select(DockerDiskUsage))).scalars())
+    hosts = await host_map(db, {s.host_id for s in summaries})
+    out = []
+    for s in summaries:
+        name = hosts[s.host_id].name if s.host_id in hosts else s.host_id
+        parts = {m: int(getattr(s, f) or 0) for m, f in DISK_PARTS.items()}
+        series = await metric_series.points(db, "docker", name, "disk_total_bytes", since, target=points)
+        out.append({"host_id": s.host_id, "host_name": name, "as_of": s.updated_at,
+                    "total": sum(parts.values()), "parts": parts,
+                    "reclaimable": int(s.images_reclaimable or 0), "points": series,
+                    "growth": growth(series)})
+    return sorted(out, key=lambda h: (-h["total"], h["host_name"].lower()))

plugins/docker/swarm_health.py

@@ -0,0 +1,57 @@
+"""The swarm's health as three figures an alert rule can watch (#5788).
+
+Every manager's agent reports the whole swarm, so the figures are swarm-wide,
+on one resource, `swarm`, and are recorded at most once every RECORD_EVERY
+whichever manager sent them: three managers then read as one series, and one
+manager going silent doesn't silence it.
+
+- swarm_services_short: services running fewer tasks than they want. A
+  service scaled to 0 wants none and is never short.
+- swarm_nodes_down: nodes whose status isn't ready, other than drained ones.
+  A drained node is one taken out of service on purpose (VDNT-Util01 always;
+  any node during its maintenance), so its going away is expected. Its host's
+  own "Agent offline" rule still covers it.
+- swarm_managers_unreachable: managers the swarm can't reach (agent 1.19.0
+  sends ManagerStatus.Reachability; before that, a manager whose status
+  isn't ready), drained ones aside for the same reason. With three managers
+  one can be lost and the swarm keeps working; two, and it can't change.
+
+A maintenance job on a swarm node holds these rules' notifications in its
+window (host_agent maintenance._swarm_window), since draining and rebooting a
+node is exactly what moves them.
+"""
+from __future__ import annotations
+
+from datetime import datetime, timedelta
+
+RESOURCE = "swarm"
+SWARM_METRICS = ("swarm_services_short", "swarm_nodes_down", "swarm_managers_unreachable")
+RECORD_EVERY = timedelta(seconds=30)
+
+_last_recorded: datetime | None = None
+
+
+def figures(swarm: dict) -> dict[str, float]:
+    """The three figures from one manager's swarm report."""
+    services = swarm.get("services") or []
+    nodes = swarm.get("nodes") or []
+    short = sum(1 for s in services
+                if int(s.get("desired") or 0) > 0 and int(s.get("running") or 0) < int(s.get("desired") or 0))
+    in_service = [n for n in nodes if (n.get("availability") or "active") != "drain"]
+    down = sum(1 for n in in_service if (n.get("status") or "unknown") != "ready")
+    managers = [n for n in in_service if n.get("role") == "manager"]
+    unreachable = sum(1 for n in managers
+                      if (n["reachability"] != "reachable" if n.get("reachability")
+                          else (n.get("status") or "unknown") != "ready"))
+    return {"swarm_services_short": float(short), "swarm_nodes_down": float(down),
+            "swarm_managers_unreachable": float(unreachable)}
+
+
+def due(now: datetime) -> bool:
+    """Whether this report should be recorded: the first in RECORD_EVERY,
+    whichever manager it came from."""
+    global _last_recorded
+    if _last_recorded is not None and now - _last_recorded < RECORD_EVERY:
+        return False
+    _last_recorded = now
+    return True

plugins/docker/swarm_nodes.py

@@ -0,0 +1,71 @@
+# plugins/docker/swarm_nodes.py
+"""The Swarm page laid out by node (#5303): one card per node with its role,
+whether it takes work, what its containers use against what the machine has,
+and the services Swarm placed on it. Model-free, so it is unit-tested without
+a database; routes.swarm() feeds it.
+
+A node with a Steward agent reports its own containers, so its services carry
+CPU and memory. A node without one is known only from the managers' placement:
+its services are listed by count, with no figures.
+"""
+from __future__ import annotations
+
+from .swarm_view import _placement
+
+
+def node_ok(n) -> bool:
+    """A node that is up and taking work."""
+    return n.status == "ready" and n.availability == "active"
+
+
+def node_cards(nodes, services, containers, *, node_host: dict, host_names: dict,
+               capacity: dict, service_state: dict) -> list[dict]:
+    """One card per node, a node in trouble first, then managers, then by name.
+
+    `nodes` and `services` are one swarm's deduplicated rows; `containers`
+    every container row (only running swarm tasks on these nodes count);
+    `node_host` maps a node id to the Steward host on it, `capacity` a host id
+    to (cores, memory bytes), `service_state` a service name to healthy,
+    degraded or down."""
+    running_on: dict[str, dict[str, dict]] = {}
+    for c in containers:
+        if not (c.node_id and c.service_name and c.status == "running"):
+            continue
+        row = running_on.setdefault(c.node_id, {}).setdefault(
+            c.service_name, {"name": c.service_name, "tasks": 0, "cpu_pct": 0.0, "mem_bytes": 0,
+                             "measured": True})
+        row["tasks"] += 1
+        row["cpu_pct"] += c.cpu_pct or 0.0
+        row["mem_bytes"] += c.mem_usage_bytes or 0
+    placed: dict[str, dict[str, int]] = {}
+    for s in services:
+        for p in _placement(s):
+            count = int(p.get("running") or 0)
+            if count > 0 and p.get("node_id"):
+                placed.setdefault(p["node_id"], {})[s.service_name] = count
+
+    cards = []
+    for n in nodes:
+        rows = dict(running_on.get(n.node_id, {}))
+        for name, count in placed.get(n.node_id, {}).items():
+            rows.setdefault(name, {"name": name, "tasks": count, "cpu_pct": None, "mem_bytes": None,
+                                   "measured": False})
+        for r in rows.values():
+            r["state"] = service_state.get(r["name"], "healthy")
+            r["cores"] = r["cpu_pct"] / 100 if r["cpu_pct"] is not None else None
+        tasks = sorted(rows.values(), key=lambda r: (r["state"] == "healthy", -(r["cores"] or 0.0),
+                                                     r["name"].lower()))
+        host_id = node_host.get(n.node_id)
+        cores, mem = capacity.get(host_id, (None, None)) if host_id else (None, None)
+        measured = [r for r in tasks if r["measured"]]
+        cards.append({
+            "node_id": n.node_id, "hostname": n.hostname or n.node_id[:12], "role": n.role,
+            "leader": n.leader, "status": n.status, "availability": n.availability, "ok": node_ok(n),
+            "host_id": host_id, "host_name": host_names.get(host_id) if host_id else None,
+            "agent": bool(running_on.get(n.node_id)) or host_id is not None,
+            "tasks": tasks, "task_count": sum(r["tasks"] for r in tasks),
+            "cores_used": sum(r["cores"] for r in measured) if measured else None,
+            "mem_used": sum(r["mem_bytes"] for r in measured) if measured else None,
+            "cores": cores, "mem": mem,
+        })
+    return sorted(cards, key=lambda c: (c["ok"], c["role"] != "manager", c["hostname"].lower()))

plugins/docker/swarm_view.py

@@ -11,12 +11,28 @@ For every swarm service we list the real replicas we have detail for, then add
 "ghost" replicas for the remaining placement count on each node (agent-less
 nodes). Non-swarm containers pass through grouped by host, unchanged.
 
+A third source, DockerSwarmService.tasks_json (agent ≥ 1.8.0), is Swarm's own
+record of each service's newest tasks. It decides which task containers are
+*superseded* — replaced by Swarm, so neither a replica nor a crash — and gives
+each service its failures across task retries and the error of its last
+failed task. Swarm restarts a task by creating a new one (a new container), so
+failures can only be counted per service, never per container.
+
 Kept model-free (no ORM imports) so it's unit-testable and safe to import in
 tests without the plugin-loader "table already defined" gotcha.
 """
 from __future__ import annotations
 
 import json
+from datetime import datetime, timedelta, timezone
+
+# Task states that mean the task ended in error: `failed` (its container exited
+# non-zero or never started) and `rejected` (Swarm could not start it at all,
+# e.g. no such image, a bad mount).
+FAILED_TASK_STATES = ("failed", "rejected")
+# Desired states of a task Swarm still means to run. Anything else (shutdown,
+# remove, complete, orphaned) is a task Swarm has finished with.
+_CURRENT_DESIRED = ("running", "ready", "")
 
 
 def _service_state(running: int, desired: int) -> str:
@@ -36,30 +52,222 @@ def _placement(service) -> list[dict]:
     return [p for p in data if isinstance(p, dict)]
 
 
-def build_swarm_services(containers, services, node_hostname: dict, host_name: dict) -> dict:
-    """Return {"services": [...], "standalone": [...]}.
+def parse_tasks(service) -> list[dict]:
+    """A service row's task records, newest first; [] when its agent sent none."""
+    raw = getattr(service, "tasks_json", None) or "[]"
+    try:
+        data = json.loads(raw)
+    except (ValueError, TypeError):
+        return []
+    return [t for t in data if isinstance(t, dict)]
 
-    services[i] = {name, mode, desired, running, state, replicas:[...]}
-      real replica  = {ghost:False, host, host_id, name, status, cpu_pct, mem_pct,
-                       health, restart_count}
-      ghost replica = {ghost:True, host, count}   # placement with no local row
-    standalone[i] = {host_id, host, containers:[...]}  # non-swarm, by host
+
+def task_time(task: dict) -> datetime | None:
+    """A task's status timestamp. Docker writes nanoseconds; Python ≥ 3.11 keeps
+    the first six digits."""
+    raw = task.get("at") or ""
+    try:
+        at = datetime.fromisoformat(raw.replace("Z", "+00:00"))
+    except ValueError:
+        return None
+    return at if at.tzinfo else at.replace(tzinfo=timezone.utc)
+
+
+def task_failures(tasks: list[dict], since: datetime) -> int:
+    """How many tasks ended in error at or after `since`."""
+    n = 0
+    for t in tasks:
+        at = task_time(t)
+        if t.get("state") in FAILED_TASK_STATES and at is not None and at >= since:
+            n += 1
+    return n
+
+
+def last_task_error(tasks: list[dict]) -> dict | None:
+    """Why the service's newest troubled task is in trouble, from Swarm itself.
+
+    The newest task that failed or was rejected, or that Swarm attached an error
+    to while still trying (a pending task with "no suitable node", say). None
+    when no retained task has a problem.
     """
+    for t in tasks:  # newest first
+        if t.get("state") in FAILED_TASK_STATES or t.get("error"):
+            return {"state": t.get("state"), "error": t.get("error") or None,
+                    "message": t.get("message") or None, "exit_code": t.get("exit_code"),
+                    "node_id": t.get("node_id") or None, "at": t.get("at") or None,
+                    "task_id": t.get("task_id") or None, "slot": t.get("slot")}
+    return None
+
+
+def task_summary(tasks: list[dict], now: datetime, crash_loop_deaths: int,
+                 node_hostname: dict) -> dict:
+    """What a service's task records say: tasks_known, failures_1h,
+    failures_24h (None without records), crash_looping, and last_error with
+    the node's hostname as `node`."""
+    known = bool(tasks)
+    failures_1h = task_failures(tasks, now - timedelta(hours=1)) if known else None
+    last_error = last_task_error(tasks)
+    if last_error is not None:
+        nid = last_error["node_id"]
+        last_error["node"] = node_hostname.get(nid) or (nid[:12] if nid else None)
+        at = task_time(last_error)
+        last_error["age_s"] = max(0, int((now - at).total_seconds())) if at else None
+    return {
+        "tasks_known": known,
+        "failures_1h": failures_1h,
+        "failures_24h": task_failures(tasks, now - timedelta(hours=24)) if known else None,
+        "crash_looping": bool(known and failures_1h >= crash_loop_deaths),
+        "last_error": last_error,
+    }
+
+
+def _oldest_failure(tasks: list[dict], since: datetime) -> datetime | None:
+    times = [at for t in tasks if t.get("state") in FAILED_TASK_STATES
+             and (at := task_time(t)) is not None and at >= since]
+    return min(times, default=None)
+
+
+def trouble_since(tasks: list[dict], state: str, crash_looping: bool,
+                  now: datetime) -> datetime | None:
+    """When a service's current trouble began, as far as its retained tasks
+    tell; None when it has none, or the tasks cannot say.
+
+    A crash loop dates from its first failure in the last hour, and a healthy
+    service that lost tasks from its first failure in the last 24h. A service
+    short of replicas dates from its troubled slots (replicated: the slot;
+    global: the node): in each slot whose newest task is not running, the
+    oldest of its newest troubled tasks (failed, rejected, or carrying an
+    error), walking back until a running task interrupts them. Swarm keeps few
+    old tasks, so a long outage reads as the oldest one retained: a lower
+    bound.
+    """
+    if crash_looping:
+        return _oldest_failure(tasks, now - timedelta(hours=1))
+    if state == "healthy":
+        return _oldest_failure(tasks, now - timedelta(hours=24))
+    by_slot: dict = {}
+    for t in tasks:  # newest first, so each slot's list is too
+        by_slot.setdefault(t.get("slot") or t.get("node_id"), []).append(t)
+    began = []
+    for slot_tasks in by_slot.values():
+        oldest = None
+        for t in slot_tasks:
+            if t.get("state") == "running":
+                break
+            if t.get("state") in FAILED_TASK_STATES or t.get("error"):
+                oldest = task_time(t) or oldest
+        if oldest is not None:
+            began.append(oldest)
+    return min(began, default=None)
+
+
+def _slot_key(c) -> str:
+    """A task container's slot within its service: "<service>.<slot>.<task>" for
+    replicated services, "<service>.<node id>.<task>" for global ones."""
+    svc, name = c.service_name, c.name or ""
+    rest = name[len(svc) + 1:] if name.startswith(svc + ".") else name
+    return rest.split(".", 1)[0]
+
+
+def superseded_containers(swarm_cs, tasks_by_id: dict[str, dict]) -> set[tuple]:
+    """(host_id, name) of task containers Swarm has replaced.
+
+    With the task on record, its desired state decides: Swarm sets a replaced
+    or failed task's desired state to shutdown. Without it (an older agent, or a
+    task older than the retained history), a container that is not running is
+    superseded when another task in the same slot is running or started later.
+    """
+    out: set[tuple] = set()
+    by_slot: dict[tuple, list] = {}
+    for c in swarm_cs:
+        by_slot.setdefault((c.service_name, _slot_key(c)), []).append(c)
+    for c in swarm_cs:
+        task = tasks_by_id.get(getattr(c, "task_id", None) or "")
+        if task is not None:
+            if task.get("desired_state") not in _CURRENT_DESIRED:
+                out.add((c.host_id, c.name))
+            continue
+        if (getattr(c, "status", "") or "").lower() == "running":
+            continue
+        mine = getattr(c, "started_at", None)
+        for other in by_slot[(c.service_name, _slot_key(c))]:
+            if other is c:
+                continue
+            theirs = getattr(other, "started_at", None)
+            if ((getattr(other, "status", "") or "").lower() == "running"
+                    or (theirs is not None and (mine is None or theirs > mine))):
+                out.add((c.host_id, c.name))
+                break
+    return out
+
+
+def _newest_per_name(services) -> dict[str, object]:
+    """Every manager reports the same cluster-global service set; keep the most
+    recently updated row per name, so a manager on an older agent (no tasks)
+    doesn't shadow one that reports them."""
+    by_name: dict[str, object] = {}
+    for s in services:
+        name = getattr(s, "service_name", None)
+        if not name:
+            continue
+        kept = by_name.get(name)
+        if kept is None:
+            by_name[name] = s
+            continue
+        a, b = getattr(s, "updated_at", None), getattr(kept, "updated_at", None)
+        if a is not None and (b is None or a > b):
+            by_name[name] = s
+    return by_name
+
+
+def replaced_tasks(containers, services) -> set[tuple]:
+    """(host_id, name) of the task containers Swarm has replaced, from the
+    container rows and the managers' service rows. What the container views use
+    to leave history out of their problem lists."""
+    swarm_cs = [c for c in containers if getattr(c, "service_name", None)]
+    tasks_by_id = {t["task_id"]: t for s in _newest_per_name(services).values()
+                   for t in parse_tasks(s) if t.get("task_id")}
+    return superseded_containers(swarm_cs, tasks_by_id)
+
+
+def build_swarm_services(containers, services, node_hostname: dict, host_name: dict,
+                         *, now: datetime | None = None, crash_loop_deaths: int = 3) -> dict:
+    """Return {"services": [...], "standalone": [...], "superseded": {(host_id, name)}}.
+
+    services[i] = {name, mode, image, desired, running, state, replicas:[...],
+                   previous_tasks, tasks_known, failures_1h, failures_24h,
+                   crash_looping, last_error}
+      real replica  = {ghost:False, host, host_id, name, status, cpu_pct, mem_pct,
+                       mem_bytes, started_at, image, health, restart_count}
+                      # current tasks only
+      ghost replica = {ghost:True, host, count}   # placement with no local row
+      previous_tasks = task containers Swarm replaced, left out of replicas
+      failures_*    = tasks that failed or were rejected in the window (None
+                      without task data); crash_looping = failures_1h ≥ the
+                      Settings → Thresholds crash-loop count
+      last_error    = last_task_error(), plus `node` (the node's hostname)
+      trouble_since = trouble_since(): when the current trouble began, or None
+    standalone[i] = {host_id, host, containers:[...]}  # non-swarm, by host
+    superseded    = the task containers Swarm replaced, so callers can skip them
+    """
+    now = now or datetime.now(timezone.utc)
     containers = list(containers)
     swarm_cs = [c for c in containers if getattr(c, "service_name", None)]
     standalone_cs = [c for c in containers if not getattr(c, "service_name", None)]
 
-    # Dedup service rows by name (every manager reports the same cluster-global
-    # set); keep the first — placement is identical across managers.
-    svc_by_name: dict[str, object] = {}
-    for s in services:
-        name = getattr(s, "service_name", None)
-        if name and name not in svc_by_name:
-            svc_by_name[name] = s
+    svc_by_name = _newest_per_name(services)
+    tasks_by_svc = {name: parse_tasks(s) for name, s in svc_by_name.items()}
+    tasks_by_id = {t["task_id"]: t for ts in tasks_by_svc.values() for t in ts
+                   if t.get("task_id")}
+    superseded = superseded_containers(swarm_cs, tasks_by_id)
 
-    # Real containers grouped by service name.
+    # Current task containers grouped by service name; replaced ones are counted.
     cs_by_service: dict[str, list] = {}
+    previous: dict[str, int] = {}
     for c in swarm_cs:
+        if (c.host_id, c.name) in superseded:
+            previous[c.service_name] = previous.get(c.service_name, 0) + 1
+            continue
         cs_by_service.setdefault(c.service_name, []).append(c)
 
     def _host_of(c):
@@ -82,6 +290,9 @@ def build_swarm_services(containers, services, node_hostname: dict, host_name: d
                 "status": getattr(c, "status", "unknown"),
                 "cpu_pct": getattr(c, "cpu_pct", None),
                 "mem_pct": getattr(c, "mem_pct", None),
+                "mem_bytes": getattr(c, "mem_usage_bytes", None),
+                "started_at": getattr(c, "started_at", None),
+                "image": getattr(c, "image", None) or None,
                 "health": getattr(c, "health", None),
                 "restart_count": getattr(c, "restart_count", 0) or 0,
             })
@@ -117,13 +328,20 @@ def build_swarm_services(containers, services, node_hostname: dict, host_name: d
         if desired is None:
             desired = len(replicas)
 
+        tasks = tasks_by_svc.get(name, [])
+        state = _service_state(int(running or 0), int(desired or 0))
+        summary = task_summary(tasks, now, crash_loop_deaths, node_hostname)
         out_services.append({
             "name": name,
             "mode": getattr(svc, "mode", "replicated") if svc is not None else "replicated",
+            "image": getattr(svc, "image", None) if svc is not None else None,
             "running": running,
             "desired": desired,
-            "state": _service_state(int(running or 0), int(desired or 0)),
+            "state": state,
             "replicas": replicas,
+            "previous_tasks": previous.get(name, 0),
+            **summary,
+            "trouble_since": trouble_since(tasks, state, summary["crash_looping"], now),
         })
 
     # Non-swarm containers grouped by host, running-first order preserved.
@@ -137,4 +355,34 @@ def build_swarm_services(containers, services, node_hostname: dict, host_name: d
         g["containers"].append(c)
     standalone = sorted(groups.values(), key=lambda g: g["host"].lower())
 
-    return {"services": out_services, "standalone": standalone}
+    return {"services": out_services, "standalone": standalone, "superseded": superseded}
+
+
+def service_trouble(s: dict) -> int | None:
+    """How bad a built service is, for sorting and for Needs attention: 0 a
+    crash loop, 1 no replica running, 2 short of replicas, 3 healthy but with
+    failed tasks in the last 24h; None when nothing is wrong."""
+    if s.get("crash_looping"):
+        return 0
+    if s.get("state") == "down":
+        return 1
+    if s.get("state") == "degraded":
+        return 2
+    if s.get("failures_24h"):
+        return 3
+    return None
+
+
+def services_problems_first(services: list[dict]) -> list[dict]:
+    """Services in trouble first, worst first, then the rest by name."""
+    return sorted(services, key=lambda s: (
+        4 if service_trouble(s) is None else service_trouble(s), s["name"].lower()))
+
+
+def failing_services(services: list[dict]) -> list[dict]:
+    """The services that belong in Needs attention: a crash loop, or short of
+    their desired replicas. A healthy service whose failed tasks Swarm has
+    already replaced is history, shown on the service, not a problem now."""
+    # A crash loop ranks 0, so test the rank against None, not for truth.
+    return [s for s in services_problems_first(services)
+            if (t := service_trouble(s)) is not None and t <= 2]

plugins/docker/templates/docker/_container_history.html

@@ -1,17 +0,0 @@
-{# docker/_container_history.html — CPU/mem sparklines for the selected range #}
-{% if have_data %}
-<div style="display:grid;grid-template-columns:repeat(auto-fit,minmax(280px,1fr));gap:1.5rem;">
-  <div>
-    <div class="section-title" style="margin-bottom:0.4rem;">CPU %</div>
-    {{ sparkline_cpu | safe }}
-  </div>
-  <div>
-    <div class="section-title" style="margin-bottom:0.4rem;">Memory %</div>
-    {{ sparkline_mem | safe }}
-  </div>
-</div>
-{% else %}
-<div style="color:var(--text-muted);font-size:0.85rem;">
-  Not enough samples in this range yet — history fills in as the agent reports.
-</div>
-{% endif %}

plugins/docker/templates/docker/_container_logs_lines.html

@@ -0,0 +1,24 @@
+{# docker/_container_logs_lines.html — log-line fragment, HTMX-polled (m79).
+   {{ l.line }} is auto-escaped by Jinja, so log content can't inject markup.
+   A service's lines (#5302) carry the task each came from. #}
+{% if lines %}
+{% for l in lines %}
+<div style="display:flex;gap:0.6rem;white-space:pre-wrap;word-break:break-word;">
+  <span style="color:var(--text-dim);flex-shrink:0;" title="{{ l.ts }}">{{ l.ts.strftime("%m-%d %H:%M:%S") }}</span>
+  {% if l.task %}<span class="tone-muted" style="flex-shrink:0;white-space:nowrap;" title="Slot and task id">{{ l.task }}</span>{% endif %}
+  <span style="flex-shrink:0;width:3.2rem;color:{% if l.stream == 'stderr' %}var(--red){% else %}var(--text-muted){% endif %};">{{ l.stream }}</span>
+  <span style="flex:1;min-width:0;{% if l.stream == 'stderr' %}color:var(--text);{% endif %}">{{ l.line }}</span>
+</div>
+{% endfor %}
+{% else %}
+<div style="color:var(--text-muted);padding:1rem 0.25rem;">
+  {% if filtered %}
+  No lines match the current filter.
+  {% else %}
+  No logs collected yet for <code>{{ name }}</code>. Lines appear within a few
+  seconds of {{ "one of its tasks" if service else "the container" }} writing to stdout/stderr — unless it's on the log
+  exclude list or log collection is turned off in
+  <a href="/settings/thresholds/">Settings → Thresholds &amp; Retention</a>.
+  {% endif %}
+</div>
+{% endif %}

plugins/docker/templates/docker/_disk_hosts.html

@@ -0,0 +1,37 @@
+{# docker/_disk_hosts.html — every Docker host's disk side by side (#5304):
+   what it uses now, how much that grew over the selected range, and a line
+   of it; the parts after. A host is linked to its own section below. #}
+{% if rows %}
+<table class="table">
+  <thead><tr>
+    <th>Host</th>
+    <th class="num">Docker uses</th>
+    <th class="num" title="Change in the total over the last {{ range_key }}, from Steward's own record of it">Change ({{ range_key }})</th>
+    <th>Over {{ range_key }}</th>
+    <th class="num">Images</th><th class="num">Volumes</th><th class="num" title="Written by containers on top of their images">Container files</th>
+    <th class="num">Build cache</th><th class="num" title="Held by images no container uses">Reclaimable</th>
+  </tr></thead>
+  <tbody>
+    {% for r in rows %}
+    {% set g = r.growth %}
+    <tr>
+      <td><a href="#host-{{ r.host_id }}">{{ r.host_name }}</a></td>
+      <td class="num"><strong>{{ r.total|bytes }}</strong></td>
+      <td class="num">
+        {%- if g is none %}<span class="tone-dim" title="Steward records this from now on; growth shows once there are two readings">—</span>
+        {%- else %}<span class="tone-{{ 'warn' if g.pct is not none and g.pct >= 10 else ('ok' if g.change < 0 else 'muted') }}"
+              title="{{ g.from|bytes }} → {{ g.to|bytes }} since {{ g.since.strftime('%Y-%m-%d %H:%M UTC') }}">{{ "+" if g.change > 0 else ("−" if g.change < 0 else "±") }}{{ (g.change|abs)|bytes }}</span>{% endif -%}
+      </td>
+      <td class="tone-info">{{ r.spark|safe }}</td>
+      <td class="num">{{ r.parts.disk_images_bytes|bytes }}</td>
+      <td class="num">{{ r.parts.disk_volumes_bytes|bytes }}</td>
+      <td class="num">{{ r.parts.disk_containers_bytes|bytes }}</td>
+      <td class="num">{{ r.parts.disk_build_cache_bytes|bytes }}</td>
+      <td class="num">{% if r.reclaimable %}<span class="tone-warn">{{ r.reclaimable|bytes }}</span>{% else %}<span class="tone-dim">—</span>{% endif %}</td>
+    </tr>
+    {% endfor %}
+  </tbody>
+</table>
+{% else %}
+<p class="tone-muted" style="margin:0;padding:1rem;">No host has reported its Docker disk yet.</p>
+{% endif %}

plugins/docker/templates/docker/_swarm_failures.html

@@ -0,0 +1,25 @@
+{#- One line on a swarm service's task failures, from Swarm's own task records
+    (agent ≥ 1.8.0): a crash loop or the failures of the last 24h, and how the
+    newest troubled task ended, in Swarm's words. Expects `s` with state,
+    crash_looping, failures_1h, failures_24h and last_error; `style` is
+    optional extra CSS for the wrapper. Renders nothing for a healthy service
+    with no recent failures. -#}
+{%- set e = s.last_error -%}
+{%- if s.crash_looping or s.failures_24h or (e and s.state != 'healthy') %}
+<div style="display:flex;flex-wrap:wrap;align-items:baseline;gap:0.2rem 0.6rem;font-size:0.78rem;color:var(--text-muted);{{ style or '' }}">
+  {% if s.crash_looping %}
+  <span style="color:var(--red);font-weight:500;">Crash-looping · {{ s.failures_1h }} tasks failed in the last hour</span>
+  {% elif s.failures_24h %}
+  <span style="color:var(--orange);">{{ s.failures_24h }} task{{ 's' if s.failures_24h != 1 }} failed in the last 24h</span>
+  {% endif %}
+  {% if e %}
+  <span title="How Swarm recorded the end of this service's newest failed task">
+    Last task {{ e.state }}:
+    <span style="color:var(--text);font-family:ui-monospace,monospace;overflow-wrap:anywhere;">{{ e.error or e.message or e.state }}</span>
+    {%- if e.exit_code %} · exit {{ e.exit_code }}{% endif %}
+    {%- if e.node %} · {{ e.node }}{% endif %}
+    {%- if e.age_s is not none %} · {{ e.age_s|duration }} ago{% endif %}
+  </span>
+  {% endif %}
+</div>
+{%- endif %}

plugins/docker/templates/docker/_usage.html

@@ -0,0 +1,28 @@
+{#- A container's CPU and memory as service.usage() describes them (#5209).
+
+    Docker counts CPU per core (100% = one core busy), so CPU reads as cores,
+    toned by its share of the host's CPUs when the host's core count is known.
+    Memory reads in bytes, against a limit only when one is actually set;
+    mem_pct is then a share of that limit, otherwise of the host's memory.
+    Each is toned by usage()'s grade when it has one: within the container's
+    own normal for the hour reads healthy, and says what is usual (#5742). -#}
+
+{% macro cores(n) -%}
+{{ "%.2f"|format(n) if n < 1 else "%.1f"|format(n) }} {{ "core" if n == 1 else "cores" }}
+{%- endmacro %}
+
+{% macro cpu(u) -%}
+{%- if u.cores is none -%}<span class="tone-dim">—</span>
+{%- else -%}
+<span class="tone-{{ u.cpu_grade or status_class(u.cpu_share, 'cpu') }}"
+      title="Docker counts CPU per core: {{ '%.0f'|format(u.cores * 100) }}% of one core{% if u.cpus %}, {{ '%.1f'|format(u.cpu_share) }}% of this host's {{ u.cpus }} core{{ 's' if u.cpus != 1 }}{% endif %}{% if u.cpu_usual %}; {{ u.cpu_usual }}{% endif %}">{{ cores(u.cores) }}</span>
+{%- endif -%}
+{%- endmacro %}
+
+{% macro mem(u) -%}
+{%- if u.mem_bytes is none -%}<span class="tone-dim">—</span>
+{%- else -%}
+<span class="tone-{{ u.mem_grade or (status_class(u.mem_pct, 'mem') if u.mem_of else 'dim') }}"
+      title="{% if u.mem_pct is not none and u.mem_of %}{{ '%.1f'|format(u.mem_pct) }}% of {{ u.mem_of }}{% if u.mem_usual %}, {{ u.mem_usual }}{% endif %}{% else %}No memory limit set{% endif %}">{{ u.mem_bytes|bytes }}{% if u.mem_limit %} / {{ u.mem_limit|bytes }}{% endif %}</span>
+{%- endif -%}
+{%- endmacro %}

plugins/docker/templates/docker/_usage_history.html

@@ -0,0 +1,23 @@
+{# docker/_usage_history.html — CPU and memory over the selected range, for a
+   service (summed across its replicas, #5302) or one container (#5310): the
+   latest and the peak, over a line that spans its half. The line is drawn
+   against at least cpu_floor / mem_floor, so idling near zero draws flat. #}
+{% from "docker/_usage.html" import cores %}
+{% if have_data %}
+<div class="history-pair">
+  <div>
+    <div class="section-title">CPU</div>
+    <div style="font-variant-numeric:tabular-nums;">{{ cores(cpu_now) }} <span class="tone-dim">now · peak {{ cores(cpu_peak) }}</span></div>
+    <div class="spark" title="Drawn from zero to its peak, or to {{ cores(cpu_floor)|striptags }} when the peak is lower">{{ spark_cpu|safe }}</div>
+  </div>
+  <div>
+    <div class="section-title">Memory</div>
+    <div style="font-variant-numeric:tabular-nums;">{{ mem_now|bytes }} <span class="tone-dim">now · peak {{ mem_peak|bytes }}</span></div>
+    <div class="spark" title="Drawn from zero to its peak, or to {{ mem_floor|bytes }} when the peak is lower">{{ spark_mem|safe }}</div>
+  </div>
+</div>
+{% else %}
+<p class="tone-muted" style="margin:0;font-size:0.85rem;">
+  Not enough samples in the last {{ range_key }} yet. History fills in as the agent reports.
+</p>
+{% endif %}

plugins/docker/templates/docker/container_detail.html

@@ -1,78 +1,98 @@
-{# docker/container_detail.html — full detail page for one container #}
+{# docker/container_detail.html — one container (#5310), shaped like a
+   service's page: a header that says whether it is well and what it is; why
+   it stopped, only while it has; what it moved since it started; its CPU and
+   memory over time (the one home for what it uses now), with its metric
+   anomalies above when it has any; its lifecycle.
+   A swarm task is titled by its service and slot, the raw name under it. #}
 {% extends "base.html" %}
-{% from "_macros.html" import crumbs %}
-{% block title %}{{ name }} — Docker — Steward{% endblock %}
-{% block breadcrumb %}{{ crumbs([("Docker", "/plugins/docker/"), (name, "")]) }}{% endblock %}
+{% from "_macros.html" import crumbs, status_badge %}
+{% block title %}{{ label or name }} — Docker — Steward{% endblock %}
+{% block breadcrumb %}
+{%- if container is not none and container.service_name -%}
+{{ crumbs([("Docker", "/plugins/docker/"), (container.service_name, "/plugins/docker/service/" ~ container.service_name), (label, "")]) }}
+{%- else -%}
+{{ crumbs([("Docker", "/plugins/docker/"), (label or name, "")]) }}
+{%- endif -%}
+{% endblock %}
 {% block content %}
 
 {% if container is none %}
-<div class="card" style="text-align:center;padding:3rem;">
-  <div style="font-weight:600;margin-bottom:0.4rem;">Container not found</div>
-  <div style="color:var(--text-muted);font-size:0.9rem;">
-    No container named <code>{{ name }}</code> is currently reported for this host.
-    It may have been removed, or the host agent hasn't reported recently.
-  </div>
+<div class="card empty-state">
+  <strong>Container not found</strong>
+  <span>No container named <code>{{ name }}</code> is reported for this host now. It may have been
+    removed, or the host agent has not reported recently. <a href="/plugins/docker/">Back to Docker</a></span>
 </div>
 {% else %}
+{% set c = container %}
 
-{# ── Header ──────────────────────────────────────────────────────────────── #}
-<div style="display:flex;align-items:center;gap:0.6rem;margin-bottom:0.35rem;flex-wrap:wrap;">
-  <span class="dot {% if container.status == 'running' %}dot-up{% elif container.status == 'paused' %}dot-warn{% else %}dot-down{% endif %}"></span>
-  <h1 class="page-title" style="margin-bottom:0;">{{ container.name }}</h1>
-  {% if container.health == 'healthy' %}<span style="font-size:0.72rem;color:var(--green);border:1px solid var(--green);border-radius:3px;padding:0.05rem 0.4rem;">healthy</span>
-  {% elif container.health == 'unhealthy' %}<span style="font-size:0.72rem;color:var(--red);border:1px solid var(--red);border-radius:3px;padding:0.05rem 0.4rem;">unhealthy</span>
-  {% elif container.health == 'starting' %}<span style="font-size:0.72rem;color:var(--orange);border:1px solid var(--orange);border-radius:3px;padding:0.05rem 0.4rem;">starting</span>{% endif %}
-</div>
-<div style="font-size:0.82rem;color:var(--text-muted);margin-bottom:1.5rem;">
-  {{ container.status }}{% if uptime %} · up {{ uptime }}{% endif %}
-  {% if host %} · on <a href="/hosts/{{ host.id }}" style="color:var(--text-muted);">{{ host.name }}</a>{% endif %}
-</div>
-
-{# ── Facts grid ──────────────────────────────────────────────────────────── #}
-{% macro fact(label, value, colour="") %}
-<div class="card" style="margin-bottom:0;">
-  <div class="section-title" style="margin-bottom:0.3rem;">{{ label }}</div>
-  <div style="font-size:0.95rem;{% if colour %}color:{{ colour }};{% endif %}font-family:ui-monospace,monospace;">{{ value }}</div>
-</div>
-{% endmacro %}
-
-<div style="display:grid;grid-template-columns:repeat(auto-fill,minmax(150px,1fr));gap:0.75rem;margin-bottom:1.5rem;">
-  {{ fact("CPU", "%.1f%%" | format(container.cpu_pct) if container.cpu_pct is not none else "—") }}
-  {{ fact("Memory", "%.1f%%" | format(container.mem_pct) if container.mem_pct is not none else "—") }}
-  {{ fact("Restarts", container.restart_count, "var(--orange)" if container.restart_count else "") }}
-  {{ fact("Last exit", (container.exit_code ~ (" (OOM)" if container.oom_killed else "")) if container.exit_code is not none else "—",
-          "var(--red)" if (container.exit_code is not none and container.exit_code != 0) else "") }}
-  {{ fact("Net in", net_rx) }}
-  {{ fact("Net out", net_tx) }}
-  {{ fact("Block read", blk_read) }}
-  {{ fact("Block write", blk_write) }}
-</div>
-
-{# ── Image / placement ───────────────────────────────────────────────────── #}
-<div class="card" style="margin-bottom:1.5rem;">
-  <div style="display:grid;grid-template-columns:max-content 1fr;gap:0.4rem 1.25rem;font-size:0.86rem;">
-    <span style="color:var(--text-muted);">Image</span>
-    <span style="font-family:ui-monospace,monospace;word-break:break-all;">{{ container.image or "—" }}</span>
-    {% if container.compose_project %}
-    <span style="color:var(--text-muted);">Compose project</span><span>{{ container.compose_project }}</span>
-    {% endif %}
-    {% if container.service_name %}
-    <span style="color:var(--text-muted);">Swarm service</span><span>{{ container.service_name }}</span>
-    {% endif %}
-    {% if container.node_id %}
-    <span style="color:var(--text-muted);">Node</span><span style="font-family:ui-monospace,monospace;">{{ container.node_id }}</span>
-    {% endif %}
-    {% if ports %}
-    <span style="color:var(--text-muted);">Ports</span>
-    <span style="font-family:ui-monospace,monospace;">{% for p in ports %}{{ p.host_port }}:{{ p.container_port }}/{{ p.protocol }}{% if not loop.last %}, {% endif %}{% endfor %}</span>
-    {% endif %}
+{# ── Header: the verdict, then what it is ────────────────────────────────── #}
+<div class="page-header">
+  <div style="min-width:0;">
+    <h1 class="page-title" style="overflow-wrap:anywhere;">{{ label }}</h1>
+    <p class="page-subtitle">
+      {% if c.status == 'running' %}{{ status_badge('ok', 'Running') }}
+      {% elif c.status == 'restarting' %}{{ status_badge('crit', 'Restarting') }}
+      {% elif c.status == 'paused' %}{{ status_badge('warn', 'Paused') }}
+      {% elif exit_failed %}{{ status_badge('crit', c.status|capitalize) }}
+      {% else %}{{ status_badge('dim', c.status|capitalize) }}{% endif %}
+      {%- if c.health == 'unhealthy' %} {{ status_badge('crit', 'Unhealthy') }}
+      {%- elif c.health == 'starting' %} {{ status_badge('pending', 'Starting') }}
+      {%- elif c.health == 'healthy' %} {{ status_badge('ok', 'Healthy') }}{% endif %}
+      {%- if uptime %} · up {{ uptime }}{% endif %}
+      {%- if host %} · on <a href="/hosts/{{ host.id }}">{{ host.name }}</a>{% endif %}
+      {%- if c.service_name %} · task of <a href="/plugins/docker/service/{{ c.service_name }}">{{ c.service_name }}</a>
+      {%- elif c.compose_project %} · compose <span class="chip">{{ c.compose_project }}</span>{% endif %}
+      {%- if c.image %} · <span class="mono" style="overflow-wrap:anywhere;" title="{{ c.image }}">{{ c.image.split("@")[0] }}</span>{% endif %}
+    </p>
+    <p class="page-subtitle tone-dim" style="margin-top:0.2rem;">
+      {%- if label != c.name %}<span class="mono" style="overflow-wrap:anywhere;">{{ c.name }}</span> · {% endif -%}
+      {%- if c.scraped_at %}reported {{ c.scraped_at|ago }}{% endif -%}
+    </p>
   </div>
+  <div class="page-actions"><a class="btn btn-sm btn-ghost" href="/plugins/docker/container/{{ host_id }}/{{ name }}/logs">Logs</a></div>
 </div>
 
-{# ── Resource history (range-toggled) ────────────────────────────────────── #}
-<div class="card" style="margin-bottom:1.5rem;">
+{# ── Why it stopped, only while it has ───────────────────────────────────── #}
+{% if c.status != 'running' and exit_failed %}
+<section class="card problem-card" aria-labelledby="ctr-exit">
+  <h2 id="ctr-exit" class="section-title" style="margin-bottom:0.4rem;">Last exit</h2>
+  <p style="margin:0;">
+    Exited <span class="tone-crit">{{ c.exit_code }}</span>{% if c.oom_killed %}: killed for running out of memory{% endif %}
+    {%- if died_at %} · {{ died_at|ago }}{% endif %}
+    · <a href="/plugins/docker/container/{{ host_id }}/{{ name }}/logs">its logs →</a>
+  </p>
+</section>
+{% endif %}
+
+{# ── Its CPU or memory out of its normal for the hour, and past anomalies (#5742) #}
+{% with back=anomaly_back, base_resource=anomaly_resource %}{% include "_anomaly_card.html" %}{% endwith %}
+
+{# ── What it is set up with, and what it moved since it started ─────────── #}
+<div class="card">
+  <dl class="kv">
+    {% if not c.service_name %}
+    {#- Swarm never restarts a task's container: it replaces the task, so a
+        task's restart count is always 0; its service's Tasks table is the
+        count that means something. #}
+    <dt>Restarts</dt><dd>{% if c.restart_count %}<span class="tone-warn">{{ c.restart_count }}</span>{% else %}0{% endif %}
+      {%- if c.exit_code is not none and c.status == 'running' and c.restart_count %} <span class="tone-dim">· last exit {{ c.exit_code }}</span>{% endif %}</dd>
+    {% endif %}
+    <dt>Memory limit</dt><dd>{% if usage.mem_limit %}{{ usage.mem_limit|bytes }}{% else %}<span class="tone-dim">none set</span>{% endif %}</dd>
+    {% if ports %}
+    <dt>Ports</dt><dd class="mono">{% for p in ports %}{{ p.host_port }}:{{ p.container_port }}/{{ p.protocol }}{% if not loop.last %}, {% endif %}{% endfor %}</dd>
+    {% endif %}
+    {% if c.node_id %}
+    <dt>Swarm node</dt><dd title="node id {{ c.node_id }}">{{ node_name or c.node_id[:12] }}</dd>
+    {% endif %}
+    <dt>Network</dt><dd>{{ net_rx }} in · {{ net_tx }} out <span class="tone-dim">since it started</span></dd>
+    <dt>Disk</dt><dd>{{ blk_read }} read · {{ blk_write }} written <span class="tone-dim">since it started</span></dd>
+  </dl>
+</div>
+
+{# ── Resource history: the one home for what it uses ────────────────────── #}
+<section class="card" aria-labelledby="ctr-history">
   <div style="display:flex;align-items:center;justify-content:space-between;gap:1rem;margin-bottom:0.75rem;flex-wrap:wrap;">
-    <h3 class="section-title" style="margin-bottom:0;">Resource history</h3>
+    <h2 id="ctr-history" class="section-title" style="margin:0;">Resource history</h2>
     {% include "_time_range.html" %}
   </div>
   <div id="container-history"
@@ -80,31 +100,30 @@
        hx-trigger="load, rangeChange from:body"
        hx-include="#time-range"
        hx-swap="innerHTML">
-    <div style="color:var(--text-muted);font-size:0.85rem;">Loading…</div>
+    <p class="field-hint">Loading…</p>
   </div>
-</div>
+</section>
 
-{# ── Lifecycle timeline ──────────────────────────────────────────────────── #}
-<div class="card">
-  <h3 class="section-title" style="margin-bottom:0.75rem;">Lifecycle</h3>
+{# ── Lifecycle, newest first ─────────────────────────────────────────────── #}
+<section class="card-flush" aria-labelledby="ctr-lifecycle">
+  <h2 id="ctr-lifecycle" class="table-caption">Lifecycle <span class="tone-muted">newest first</span></h2>
   {% if events %}
-  <div style="display:grid;gap:0.5rem;">
-    {% for e in events %}
-    <div style="display:flex;align-items:baseline;gap:0.6rem;font-size:0.85rem;">
-      <span style="color:{{ e.colour }};width:1rem;text-align:center;flex-shrink:0;">{{ e.glyph }}</span>
-      <span style="font-weight:500;width:6.5rem;flex-shrink:0;">{{ e.event }}</span>
-      <span style="color:var(--text-muted);flex:1;min-width:0;">{{ e.detail or "" }}</span>
-      <span style="color:var(--text-dim);font-size:0.78rem;white-space:nowrap;" title="{{ e.at }}">{{ e.at.strftime("%Y-%m-%d %H:%M") }}</span>
-    </div>
-    {% endfor %}
-  </div>
+  <table class="table">
+    <thead><tr><th>Event</th><th>Detail</th><th class="num">When</th></tr></thead>
+    <tbody>
+      {% for e in events %}
+      <tr>
+        <td><span class="tone-{{ e.tone }}" aria-hidden="true">{{ e.glyph }}</span> {{ e.event|replace("_", " ") }}</td>
+        <td class="tone-muted" style="overflow-wrap:anywhere;">{{ e.detail or "" }}</td>
+        <td class="num">{{ e.at|ago }}</td>
+      </tr>
+      {% endfor %}
+    </tbody>
+  </table>
   {% else %}
-  <div style="color:var(--text-muted);font-size:0.85rem;">
-    No lifecycle events recorded yet. Start/stop/health changes appear here as the
-    agent reports them over time.
-  </div>
+  <p class="tone-muted" style="margin:0;padding:1rem;">No lifecycle events recorded yet. Starts, stops and health changes appear here as the agent reports them.</p>
   {% endif %}
-</div>
+</section>
 
 {% endif %}
 {% endblock %}

plugins/docker/templates/docker/container_logs.html

@@ -0,0 +1,51 @@
+{# docker/container_logs.html — per-container log viewer (m79) #}
+{% extends "base.html" %}
+{% from "_macros.html" import crumbs %}
+{% block title %}Logs — {{ name }} — Docker — Steward{% endblock %}
+{% block breadcrumb %}{{ crumbs([
+  ("Docker", "/plugins/docker/"),
+  (name, "/plugins/docker/container/" ~ host_id ~ "/" ~ name),
+  ("Logs", "")]) }}{% endblock %}
+{% block content %}
+
+<div style="display:flex;align-items:baseline;gap:0.6rem;margin-bottom:0.35rem;flex-wrap:wrap;">
+  <h1 class="page-title" style="margin-bottom:0;">{{ name }}</h1>
+  <span style="font-size:0.9rem;color:var(--text-muted);">logs</span>
+</div>
+<div style="font-size:0.82rem;color:var(--text-muted);margin-bottom:1rem;">
+  Recent lines (newest first) collected by the host agent{% if host %} on
+  <a href="/hosts/{{ host.id }}" style="color:var(--text-muted);">{{ host.name }}</a>{% endif %} —
+  updated every few seconds.
+  <a href="/plugins/docker/container/{{ host_id }}/{{ name }}">← back to container</a>
+</div>
+
+{# ── Controls: stream filter + text search (drive the fragment via HTMX) ───── #}
+<form id="log-controls" onsubmit="return false;"
+      style="display:flex;gap:0.6rem;align-items:center;flex-wrap:wrap;margin-bottom:0.6rem;">
+  <label style="font-size:0.8rem;color:var(--text-muted);display:flex;align-items:center;gap:0.35rem;">
+    Stream
+    <select name="stream" style="padding:0.25rem 0.4rem;">
+      <option value="all">all</option>
+      <option value="stdout">stdout</option>
+      <option value="stderr">stderr</option>
+    </select>
+  </label>
+  <input type="search" name="q" placeholder="Filter lines…" autocomplete="off"
+         aria-label="Filter log lines"
+         style="flex:1;min-width:180px;padding:0.3rem 0.5rem;">
+</form>
+
+<div class="card-flush">
+  <div id="log-lines"
+       hx-get="/plugins/docker/container/{{ host_id }}/{{ name }}/logs/lines"
+       hx-trigger="load, every 5s, change from:#log-controls, keyup changed delay:400ms from:#log-controls"
+       hx-include="#log-controls"
+       hx-swap="innerHTML"
+       role="log" aria-live="polite" tabindex="0"
+       style="max-height:70vh;overflow:auto;padding:0.5rem 0.75rem;
+              font-family:ui-monospace,monospace;font-size:0.8rem;line-height:1.5;">
+    <div style="color:var(--text-muted);">Loading…</div>
+  </div>
+</div>
+
+{% endblock %}

plugins/docker/templates/docker/disk.html

@@ -1,83 +1,124 @@
-{# docker/disk.html — image/disk usage: reclaimable space, per-image sizes #}
+{# docker/disk.html — what Docker keeps on each host's disk (#5304, #5305):
+   the hosts side by side with how much each grew, then each host's parts,
+   its cleanup actions, and its images folded under their repository. #}
 {% extends "base.html" %}
-{% from "_macros.html" import crumbs %}
+{% from "_macros.html" import crumbs, page_header, stat_tile, status_badge %}
 {% block title %}Disk — Docker — Steward{% endblock %}
-{% block breadcrumb %}{{ crumbs([("Docker", "/plugins/docker/"), ("Image & disk usage", "")]) }}{% endblock %}
+{% block breadcrumb %}{{ crumbs([("Docker", "/plugins/docker/"), ("Disk", "")]) }}{% endblock %}
 {% block content %}
 
-<h1 class="page-title" style="margin-bottom:0.4rem;">Image &amp; disk usage</h1>
-<p style="font-size:0.82rem;color:var(--text-muted);margin-bottom:1.5rem;">
-  Reclaimable = space held by images no container references. Cleanup actions
-  (prune) arrive in a later release — these are read-only figures for now.
-</p>
+{% call page_header("Docker disk", "What Docker keeps on each host's disk, from docker system df, and how it has grown.") %}
+  {% include "_time_range.html" %}
+{% endcall %}
 
 {% if host_groups %}
+<section class="card-flush" aria-labelledby="disk-hosts">
+  <h2 id="disk-hosts" class="table-caption">By host <span class="tone-muted">most space first</span></h2>
+  <div hx-get="/plugins/docker/disk/hosts" hx-trigger="load, rangeChange from:body"
+       hx-include="#time-range" hx-swap="innerHTML">
+    <p class="field-hint" style="padding:1rem;margin:0;">Loading…</p>
+  </div>
+</section>
+
 {% for g in host_groups %}
-<div style="margin-bottom:2rem;">
-  <div style="font-weight:600;font-size:0.95rem;margin-bottom:0.75rem;">{{ g.host_name }}</div>
+{% set s = g.summary %}
+<section id="host-{{ g.host_id }}" class="disk-host" aria-labelledby="disk-host-{{ g.host_id }}">
+  <h2 id="disk-host-{{ g.host_id }}" class="disk-host-head">
+    <a href="/hosts/{{ g.host_id }}">{{ g.host_name }}</a>
+    <span class="tone-muted">{{ g.total|bytes }}</span>
+    {% if g.as_of %}<span class="field-hint">as of {{ g.as_of|ago }}</span>{% endif %}
+  </h2>
 
-  {# ── Summary stats ────────────────────────────────────────────────────── #}
-  <div style="display:grid;grid-template-columns:repeat(auto-fill,minmax(150px,1fr));gap:0.75rem;margin-bottom:1rem;">
-    <div class="card" style="margin-bottom:0;">
-      <div class="section-title" style="margin-bottom:0.3rem;">Reclaimable</div>
-      <span class="stat-val" style="color:{% if g.summary.images_reclaimable != '0 B' %}var(--orange){% else %}var(--green){% endif %};">{{ g.summary.images_reclaimable }}</span>
-    </div>
-    <div class="card" style="margin-bottom:0;">
-      <div class="section-title" style="margin-bottom:0.3rem;">Images</div>
-      <span class="stat-val">{{ g.summary.images_size }}</span>
-      <div style="font-size:0.74rem;color:var(--text-muted);">{{ g.summary.images_active }}/{{ g.summary.images_total }} in use</div>
-    </div>
-    <div class="card" style="margin-bottom:0;">
-      <div class="section-title" style="margin-bottom:0.3rem;">Stopped</div>
-      <span class="stat-val" style="{% if g.stopped %}color:var(--text-muted){% endif %};">{{ g.stopped }}</span>
-    </div>
-    <div class="card" style="margin-bottom:0;">
-      <div class="section-title" style="margin-bottom:0.3rem;">Writable layers</div>
-      <span class="stat-val">{{ g.summary.containers_size }}</span>
-      <div style="font-size:0.74rem;color:var(--text-muted);">{{ g.summary.containers_count }} containers</div>
-    </div>
-    <div class="card" style="margin-bottom:0;">
-      <div class="section-title" style="margin-bottom:0.3rem;">Volumes</div>
-      <span class="stat-val">{{ g.summary.volumes_size }}</span>
-      <div style="font-size:0.74rem;color:var(--text-muted);">{{ g.summary.volumes_count }} volumes</div>
-    </div>
-    <div class="card" style="margin-bottom:0;">
-      <div class="section-title" style="margin-bottom:0.3rem;">Build cache</div>
-      <span class="stat-val">{{ g.summary.build_cache_size }}</span>
-    </div>
+  <div class="stat-strip">
+    {{ stat_tile("Reclaimable", s.images_reclaimable|bytes, state="warn" if s.images_reclaimable else "ok", hint="held by images no container uses") }}
+    {{ stat_tile("Images", s.images_size|bytes, hint=s.images_active ~ " of " ~ s.images_total ~ " in use") }}
+    {{ stat_tile("Container files", s.containers_size|bytes, hint="written by " ~ s.containers_count ~ " containers on top of their images") }}
+    {{ stat_tile("Volumes", s.volumes_size|bytes, hint=s.volumes_count ~ " volumes; never pruned here") }}
+    {{ stat_tile("Build cache", s.build_cache_size|bytes, hint="left by docker build; safe to clear") }}
+    {{ stat_tile("Stopped", g.stopped, state="dim" if not g.stopped else none, hint="containers not running") }}
   </div>
 
-  {# ── Per-image table ──────────────────────────────────────────────────── #}
-  <div class="card-flush">
-    <table class="table">
-      <thead>
-        <tr><th>Image</th><th>Size</th><th>Shared</th><th>Containers</th></tr>
-      </thead>
-      <tbody>
-        {% for im in g.images %}
+  {# ── Cleanup actions (admin only; audited Ansible prune run) ───────────── #}
+  {% if session.user_role == 'admin' %}
+  <div class="disk-actions">
+    {% if ansible_available and g.has_target %}
+    <form method="post" action="/plugins/docker/disk/{{ g.host_id }}/prune"
+          data-confirm="Remove all STOPPED containers on {{ g.host_name }}? This cannot be undone.">{{ csrf_field() }}
+      <input type="hidden" name="target" value="containers">
+      <button type="submit" class="btn btn-sm">Prune stopped containers</button>
+    </form>
+    <form method="post" action="/plugins/docker/disk/{{ g.host_id }}/prune"
+          data-confirm="Remove ALL unused images on {{ g.host_name }} (docker image prune -a)? Any image not used by a container is deleted.">{{ csrf_field() }}
+      <input type="hidden" name="target" value="images">
+      <button type="submit" class="btn btn-sm">Prune unused images</button>
+    </form>
+    <form method="post" action="/plugins/docker/disk/{{ g.host_id }}/prune"
+          data-confirm="Run a full system prune on {{ g.host_name }}? Removes stopped containers, unused networks, untagged images nothing uses, and the build cache.">{{ csrf_field() }}
+      <input type="hidden" name="target" value="system">
+      <button type="submit" class="btn btn-sm btn-danger">System prune…</button>
+    </form>
+    <span class="field-hint">Each runs an audited Ansible playbook on {{ g.host_name }}; reclaimed space shows on the next agent sample.</span>
+    {% elif not ansible_available %}
+    <span class="field-hint">Cleanup needs the Ansible runner (currently unavailable).</span>
+    {% else %}
+    <span class="field-hint">Link an <a href="/hosts/{{ g.host_id }}">Ansible target</a> to this host to enable prune actions.</span>
+    {% endif %}
+  </div>
+  {% endif %}
+
+  {# ── Images by repository, versions folded under it (#5305) ─────────── #}
+  <section class="card-flush" aria-label="Images on {{ g.host_name }}">
+    <h3 class="table-caption">Images by repository
+      <span class="tone-muted">{% if g.images_capped %}the {{ g.images_listed }} largest of {{ s.images_total }} images{% else %}{{ g.images_listed }} images{% endif %}</span></h3>
+    {% if g.repos %}
+    <table class="table image-repos">
+      <thead><tr>
+        <th>Repository</th>
+        <th class="num">Versions</th>
+        <th class="num" title="Each version's size less the layers it shares with other images, added up: what removing them all would free, roughly">Own size</th>
+        <th class="num">Used by</th>
+        <th class="num" title="Own size of the versions no container uses">Reclaimable</th>
+      </tr></thead>
+      {% for r in g.repos %}
+      <tbody class="image-repo">
         <tr>
-          <td style="font-size:0.84rem;font-family:ui-monospace,monospace;max-width:340px;white-space:nowrap;overflow:hidden;text-overflow:ellipsis;">
-            {{ im.repo_tag }}
-            {% if im.reclaimable %}<span title="not referenced by any container" style="font-size:0.68rem;color:var(--orange);border:1px solid var(--orange);border-radius:3px;padding:0 0.3rem;margin-left:0.4rem;">reclaimable</span>{% endif %}
+          <td class="cell-name">
+            <details>
+              <summary title="{{ r.repo or 'untagged images' }}">{% if r.repo %}<span class="mono">{{ r.repo }}</span>{% else %}<span class="tone-muted">untagged</span>{% endif %}</summary>
+              <table class="table image-versions">
+                <thead><tr><th>Tag</th><th>ID</th><th class="num">Size</th><th class="num" title="Layers in common with other images: stored once">Shared</th><th class="num">Used by</th></tr></thead>
+                <tbody>
+                  {% for v in r.versions %}
+                  <tr>
+                    <td class="mono">{{ v.tag or "—" }}</td>
+                    <td class="mono tone-muted">{{ (v.image_id or "")|replace("sha256:", "")|truncate(12, true, "") }}</td>
+                    <td class="num">{{ v.size|bytes }}</td>
+                    <td class="num tone-muted">{{ v.shared|bytes }}</td>
+                    <td class="num">{% if v.containers %}{{ v.containers }} container{{ "s" if v.containers != 1 }}{% else %}<span class="tone-warn">nothing</span>{% endif %}</td>
+                  </tr>
+                  {% endfor %}
+                </tbody>
+              </table>
+            </details>
           </td>
-          <td style="font-family:ui-monospace,monospace;font-size:0.86rem;">{{ im.size }}</td>
-          <td style="font-family:ui-monospace,monospace;font-size:0.86rem;color:var(--text-muted);">{{ im.shared }}</td>
-          <td style="font-size:0.86rem;color:{% if im.containers %}var(--text){% else %}var(--text-muted){% endif %};">{{ im.containers }}</td>
+          <td class="num">{{ r.versions|length }}{% if r.unused %} <span class="tone-dim">({{ r.unused }} unused)</span>{% endif %}</td>
+          <td class="num">{{ r.own|bytes }}</td>
+          <td class="num">{% if r.containers %}{{ r.containers }} container{{ "s" if r.containers != 1 }}{% else %}<span class="tone-warn">nothing</span>{% endif %}</td>
+          <td class="num">{% if r.reclaimable %}<span class="tone-warn">{{ r.reclaimable|bytes }}</span>{% else %}<span class="tone-dim">—</span>{% endif %}</td>
         </tr>
-        {% else %}
-        <tr><td colspan="4" style="color:var(--text-muted);font-size:0.85rem;text-align:center;padding:1.5rem;">No images reported.</td></tr>
-        {% endfor %}
       </tbody>
+      {% endfor %}
     </table>
-  </div>
-</div>
+    {% else %}
+    <p class="tone-muted" style="margin:0;padding:1rem;">No images reported.</p>
+    {% endif %}
+  </section>
+</section>
 {% endfor %}
 {% else %}
-<div class="card" style="text-align:center;padding:3rem;">
-  <div style="color:var(--text-muted);font-size:0.9rem;">
-    No disk usage reported yet. The host agent reports image/disk usage from
-    <code>docker system df</code> on hosts running Docker.
-  </div>
+<div class="card empty-state">
+  No disk usage reported yet. The host agent reports what Docker keeps on disk,
+  from <code>docker system df</code>, on hosts running Docker.
 </div>
 {% endif %}
 {% endblock %}

plugins/docker/templates/docker/host_panel.html

@@ -1,22 +1,76 @@
-{# docker/host_panel.html — per-host Docker fragment embedded on the Hosts hub #}
+{# docker/host_panel.html — per-host Docker fragment embedded on the Hosts hub:
+   problems first with what is wrong and since when, then the busiest running
+   containers; the rest of the running and the stopped fold away so the card
+   stays short (#5276). Tasks Swarm has replaced are left out (counted in the
+   header). #}
+{% from "_macros.html" import status_badge, status_dot %}
+{% from "docker/_usage.html" import cpu, mem %}
+{% macro container_table(rows) %}
+<table class="table">
+  <colgroup><col><col style="width:7rem"><col style="width:9rem"></colgroup>
+  <thead>
+    <tr>
+      <th>Container</th>
+      <th title="Docker counts CPU per core: 1 core = one core fully busy">CPU</th>
+      <th title="Used, and the limit when one is set">Memory</th>
+    </tr>
+  </thead>
+  <tbody>
+      {% for r in rows %}
+      {% set c = r.c %}
+      <tr>
+        <td>
+          <div style="display:flex;align-items:center;gap:0.5rem;min-width:0;">
+            {{ status_dot('crit' if r.problems else ('up' if c.status == 'running' else 'inactive'), c.status|capitalize) }}
+            <div style="min-width:0;">
+              <a href="/plugins/docker/container/{{ c.host_id }}/{{ c.name }}" title="{{ c.name }} · {{ c.image }}" style="font-weight:500;color:inherit;text-decoration:none;overflow-wrap:anywhere;">{{ r.label }}</a>
+              {% if r.problems %}
+              {% set worst = r.problems[0] %}
+              <div style="font-size:0.75rem;margin-top:0.1rem;">
+                {{ status_badge('crit' if severity[worst] == 'critical' else 'warn', worst|replace('_', ' ')|capitalize) }}
+                <span style="color:var(--text-muted);">
+                  {%- for p in r.problems %} {{ says[p] }}{{ " and" if not loop.last }}{% endfor %}
+                  {%- if c.exit_code %} · exit {{ c.exit_code }}{% if c.oom_killed %} (OOM){% endif %}{% endif %}
+                  {%- if r.since %} · since {{ r.since|ago }}{% endif -%}
+                </span>
+              </div>
+              {% else %}
+              <div style="font-size:0.72rem;color:var(--text-muted);">
+                {{ c.status }}{% if r.uptime %} · up {{ r.uptime }}{% endif %}
+                {%- if c.status != 'running' and c.exit_code %} · exit {{ c.exit_code }}{% endif %}
+              </div>
+              {% endif %}
+            </div>
+          </div>
+        </td>
+        <td style="font-family:ui-monospace,monospace;font-size:0.85rem;white-space:nowrap;">{% if c.status == 'running' %}{{ cpu(r.usage) }}{% else %}<span class="tone-dim">—</span>{% endif %}</td>
+        <td style="font-family:ui-monospace,monospace;font-size:0.85rem;white-space:nowrap;">{% if c.status == 'running' %}{{ mem(r.usage) }}{% else %}<span class="tone-dim">—</span>{% endif %}</td>
+      </tr>
+      {% endfor %}
+  </tbody>
+</table>
+{% endmacro %}
 <div class="card">
-  <div style="display:flex;align-items:baseline;justify-content:space-between;gap:0.6rem;margin-bottom:0.6rem;">
+  <div style="display:flex;align-items:baseline;justify-content:space-between;gap:0.6rem;margin-bottom:0.6rem;flex-wrap:wrap;">
     <h3 class="section-title" style="margin-bottom:0;">Docker</h3>
     <span style="font-size:0.78rem;color:var(--text-muted);">
+      {% if problems %}<span class="tone-crit">{{ problems }} with problems</span> · {% endif %}
       {{ running }} running{% if stopped %} · {{ stopped }} stopped{% endif %}
+      {% if replaced %}<span title="Old swarm tasks Swarm has replaced with new ones: history, not containers that need anything"> · {{ replaced }} replaced task{{ "s" if replaced != 1 }} hidden</span>{% endif %}
       <a href="/plugins/docker/" style="margin-left:0.6rem;color:var(--text-muted);">All →</a>
     </span>
   </div>
-  <div style="display:grid;gap:2px;">
-    {% for c in containers %}
-    <div style="display:flex;align-items:center;gap:0.5rem;font-size:0.85rem;padding:0.2rem 0;">
-      <span class="dot {% if c.status == 'running' %}dot-up{% elif c.status == 'paused' %}dot-warn{% else %}dot-down{% endif %}"></span>
-      <a href="/plugins/docker/container/{{ c.host_id }}/{{ c.name }}" style="flex:1;min-width:0;white-space:nowrap;overflow:hidden;text-overflow:ellipsis;color:inherit;text-decoration:none;" title="{{ c.image }}">{{ c.name }}</a>
-      <span style="font-size:0.74rem;color:var(--text-muted);font-family:ui-monospace,monospace;flex-shrink:0;">
-        {% if c.cpu_pct is not none %}{{ "%.1f" | format(c.cpu_pct) }}%{% endif %}
-        {% if c.mem_pct is not none %} · {{ "%.1f" | format(c.mem_pct) }}%{% endif %}
-      </span>
-    </div>
-    {% endfor %}
-  </div>
+  {{ container_table(rows) }}
+  {% if folded %}
+  <details style="margin-top:0.5rem;">
+    <summary style="cursor:pointer;font-size:0.82rem;color:var(--text-muted);">
+      {%- set more_running = folded|selectattr("c.status", "equalto", "running")|list|length -%}
+      {%- set more_stopped = folded|length - more_running -%}
+      {%- if more_running %}+{{ more_running }} more running{% endif -%}
+      {%- if more_running and more_stopped %} · {% endif -%}
+      {%- if more_stopped %}{{ more_stopped }} stopped{% endif -%}
+    </summary>
+    {{ container_table(folded) }}
+  </details>
+  {% endif %}
 </div>

plugins/docker/templates/docker/index.html

@@ -1,24 +1,81 @@
 {% extends "base.html" %}
+{% from "_macros.html" import page_header %}
+{# The Docker page. The filter bar sits outside the polled fragment, so a poll
+   never resets what is being typed; the fragment reads the bar on every poll,
+   and a change re-fetches it and puts the filters in the URL (#5299). #}
 {% block title %}Docker — Steward{% endblock %}
 {% block content %}
-<div style="display:flex;align-items:center;justify-content:space-between;margin-bottom:1.5rem;gap:1rem;flex-wrap:wrap;">
-  <div style="display:flex;align-items:baseline;gap:1rem;">
-    <h1 class="page-title" style="margin-bottom:0;">Docker</h1>
-    {% if has_swarm %}
-    <a href="/plugins/docker/swarm" style="font-size:0.85rem;color:var(--accent);text-decoration:none;">Swarm →</a>
-    {% endif %}
-    {% if has_disk %}
-    <a href="/plugins/docker/disk" style="font-size:0.85rem;color:var(--accent);text-decoration:none;">Disk →</a>
-    {% endif %}
-  </div>
-  {% include "_time_range.html" %}
-</div>
+{% call page_header("Docker", "What needs attention first, then every service and container.") %}
+  {% if has_swarm %}<a class="btn btn-sm btn-ghost" href="/plugins/docker/swarm">Swarm</a>{% endif %}
+  {% if has_disk %}<a class="btn btn-sm btn-ghost" href="/plugins/docker/disk">Disk</a>{% endif %}
+{% endcall %}
+
+<form id="docker-filters" class="filter-bar" method="get" action="/plugins/docker/" role="search">
+  <input type="search" id="docker-q" name="q" value="{{ f.q }}" class="input"
+         placeholder="Search name, image or stack" aria-label="Search services and containers" autocomplete="off">
+  {% if choices.stacks %}
+  <select name="stack" aria-label="Stack">
+    <option value="">All stacks</option>
+    {% for s in choices.stacks %}<option value="{{ s }}"{% if f.stack == s %} selected{% endif %}>{{ s }}</option>{% endfor %}
+    <option value="{{ no_stack }}"{% if f.stack == no_stack %} selected{% endif %}>No stack</option>
+  </select>
+  {% endif %}
+  {% if choices.hosts|length > 1 %}
+  <select name="host" aria-label="Host or node">
+    <option value="">All hosts</option>
+    {% for h in choices.hosts %}<option value="{{ h }}"{% if f.host == h %} selected{% endif %}>{{ h }}</option>{% endfor %}
+  </select>
+  {% endif %}
+  <select name="state" aria-label="Status">
+    <option value="">Any status</option>
+    {% for key, label in states.items() %}<option value="{{ key }}"{% if f.state == key %} selected{% endif %}>{{ label }}</option>{% endfor %}
+  </select>
+  {% if f.sort != 'status' %}<input type="hidden" name="sort" value="{{ f.sort }}"><input type="hidden" name="dir" value="{{ f.dir }}">{% endif %}
+  <noscript><button type="submit" class="btn btn-sm">Filter</button></noscript>
+</form>
 
 <div id="docker-rows"
      hx-get="/plugins/docker/rows"
-     hx-trigger="load, rangeChange from:body"
-     hx-include="#time-range"
+     hx-include="#docker-filters"
+     hx-trigger="load, every {{ poll_interval }}s, change from:#docker-filters, input changed delay:250ms from:#docker-q, search from:#docker-q"
      hx-swap="innerHTML">
-  <div style="color:var(--text-muted);font-size:0.9rem;padding:2rem;">Loading...</div>
+  <div class="card"><p class="field-hint">Loading…</p></div>
 </div>
+{% if disk %}
+{# What Docker keeps on each host's disk, and its growth over a week (#5306). #}
+<p class="docker-disk-line">
+  <a href="/plugins/docker/disk" class="section-title">Disk</a>
+  {% for d in disk %}
+  <span class="docker-disk-host"><a href="/plugins/docker/disk#host-{{ d.host_id }}">{{ d.host_name }}</a>
+    {{ d.total|bytes }}
+    {%- if d.growth and d.growth.change %} <span class="tone-{{ 'warn' if d.growth.pct is not none and d.growth.pct >= 10 else 'muted' }}"
+      title="Since {{ d.growth.since.strftime('%Y-%m-%d %H:%M UTC') }}">{{ "+" if d.growth.change > 0 else "−" }}{{ (d.growth.change|abs)|bytes }} in 7d</span>{% endif -%}
+    {%- if d.reclaimable %} <span class="tone-dim">· {{ d.reclaimable|bytes }} reclaimable</span>{% endif %}</span>
+  {% endfor %}
+</p>
+{% endif %}
+<script>
+(function () {
+  var rows = document.getElementById('docker-rows');
+  var form = document.getElementById('docker-filters');
+  // The filters live in the URL too, so a filtered view can be bookmarked and
+  // Back returns to it.
+  function keepInUrl() {
+    var q = new URLSearchParams(new FormData(form));
+    Array.from(q.keys()).forEach(function (k) { if (!q.get(k)) q.delete(k); });
+    var s = q.toString();
+    history.replaceState(null, '', location.pathname + (s ? '?' + s : '') + location.hash);
+  }
+  form.addEventListener('change', keepInUrl);
+  form.addEventListener('input', keepInUrl);
+  // A link to a section (the dashboard's "failed (24h)" → #failures) lands
+  // before the rows exist; scroll to it once, when they first arrive.
+  if (!location.hash || !rows) return;
+  rows.addEventListener('htmx:afterSwap', function once() {
+    rows.removeEventListener('htmx:afterSwap', once);
+    var target = document.getElementById(location.hash.slice(1));
+    if (target) target.scrollIntoView({ block: 'start' });
+  });
+})();
+</script>
 {% endblock %}

plugins/docker/templates/docker/rows.html

@@ -1,174 +1,228 @@
-{# docker/rows.html — HTMX fragment for the Docker main page, grouped by host #}
+{# docker/rows.html — HTMX fragment for the Docker page (#5298–#5301):
+   a verdict line; what needs
+   attention (service.needs_attention, the ranking the MCP tools use); hosts
+   whose Docker is not answering; what failed in 24h, each with its logs; then
+   the swarm services and the standalone containers as two tables that share
+   the filter bar's search, filters and sort (overview.py). #}
+{% from "_macros.html" import status_badge, status_dot, sort_th %}
+{% from "docker/_usage.html" import cpu, mem, cores %}
+{% macro th(key, label, cls=none, title=none) -%}
+{{ sort_th(label, sort_url(key), active=f.sort == key, desc=f.dir == 'desc', cls=cls, title=title) }}
+{%- endmacro %}
+{% macro up(row) -%}
+{%- if row.started_at %}<span title="Started {{ row.started_at.strftime('%Y-%m-%d %H:%M UTC') }}">{{ row.started_at|ago|striptags|replace(" ago", "") }}</span>
+{%- else %}<span class="tone-dim">—</span>{% endif -%}
+{%- endmacro %}
 
-{# ── Summary strip ─────────────────────────────────────────────────────────── #}
-<div style="display:grid;grid-template-columns:repeat(auto-fill,minmax(140px,1fr));gap:0.75rem;margin-bottom:1.5rem;">
-  <div class="card" style="margin-bottom:0;">
-    <div class="section-title" style="margin-bottom:0.35rem;">Running</div>
-    <span class="stat-val" style="color:var(--green);">{{ running }}</span>
-  </div>
-  <div class="card" style="margin-bottom:0;">
-    <div class="section-title" style="margin-bottom:0.35rem;">Stopped</div>
-    <span class="stat-val" style="{% if stopped %}color:var(--text-muted){% endif %};">{{ stopped }}</span>
-  </div>
-  <div class="card" style="margin-bottom:0;">
-    <div class="section-title" style="margin-bottom:0.35rem;">Total</div>
-    <span class="stat-val">{{ total }}</span>
-  </div>
-  {% if swarm_services %}
-  <div class="card" style="margin-bottom:0;">
-    <div class="section-title" style="margin-bottom:0.35rem;">Services</div>
-    <span class="stat-val">{{ swarm_services | length }}</span>
-  </div>
-  {% endif %}
-  <div class="card" style="margin-bottom:0;">
-    <div class="section-title" style="margin-bottom:0.35rem;">Hosts</div>
-    <span class="stat-val">{{ host_groups | length }}</span>
-  </div>
-</div>
+{# ── Verdict ──────────────────────────────────────────────────────────────── #}
+{% set problems = attention|length + failing_services|length %}
+<p class="page-verdict">
+  {% if problems %}<strong class="tone-crit">{{ problems }} need{{ "s" if problems == 1 }} attention</strong>
+  {% else %}<strong class="tone-ok">Nothing needs attention</strong>{% endif %}
+  {%- if services_total %} · <span class="tone-{{ 'ok' if services_healthy == services_total else 'warn' }}">{{ services_healthy }}/{{ services_total }}</span> services healthy{% endif %}
+  {%- if containers_total %} · {{ running }} container{{ "s" if running != 1 }} running{% if stopped %}, {{ stopped }} stopped{% endif %}{% if services_total %} outside services{% endif %}{% endif %}
+  {%- if docker_hosts %} · {{ docker_hosts }} Docker host{{ "s" if docker_hosts != 1 }}{% endif %}
+  {%- if stale %} · <span class="tone-warn" title="No recent report from their agent: what is shown may no longer be true">{{ stale }} stale</span>{% endif %}
+</p>
 
-{# ── Swarm services: collapsed per-service, each replica labelled with its host.
-   Ghost replicas (tasks on nodes with no Steward agent) come from the managers'
-   placement data so the picture is cluster-complete. ──────────────────────── #}
-{% if swarm_services %}
-<div class="card-flush" style="margin-bottom:1.5rem;">
-  <div style="display:flex;align-items:baseline;gap:0.6rem;padding:0.5rem 0.75rem;border-bottom:1px solid var(--border);">
-    <span style="font-weight:600;font-size:0.95rem;">Swarm services</span>
-    <a href="/plugins/docker/swarm" style="font-size:0.78rem;color:var(--text-muted);">Topology →</a>
-  </div>
-  <div style="padding:0.5rem 0.75rem;display:grid;gap:0.85rem;">
-    {% for s in swarm_services %}
-    <div>
-      <div style="display:flex;align-items:center;gap:0.55rem;flex-wrap:wrap;margin-bottom:0.35rem;">
-        <span class="dot {% if s.state == 'healthy' %}dot-up{% elif s.state == 'degraded' %}dot-warn{% else %}dot-down{% endif %}"></span>
-        <span style="font-weight:600;font-size:0.9rem;">{{ s.name }}</span>
-        <span style="font-size:0.72rem;color:var(--text-dim);">{{ s.mode }}</span>
-        <span style="font-size:0.78rem;color:{% if s.state == 'healthy' %}var(--green){% elif s.state == 'degraded' %}var(--orange){% else %}var(--red){% endif %};font-variant-numeric:tabular-nums;">
-          {{ s.running }}/{{ s.desired }} running
-        </span>
-      </div>
-      <div style="display:grid;grid-template-columns:repeat(auto-fill,minmax(210px,1fr));gap:0.4rem;padding-left:1.1rem;">
-        {% for r in s.replicas %}
-        {% if r.ghost %}
-        <div style="display:flex;align-items:center;gap:0.4rem;font-size:0.8rem;background:var(--bg-elevated);border:1px dashed var(--border-mid);border-radius:5px;padding:0.3rem 0.55rem;" title="Running on a node with no Steward agent — detail unavailable">
-          <span class="dot dot-warn" style="opacity:0.6;"></span>
-          <span style="font-weight:500;">{{ r.host }}</span>
-          <span style="color:var(--text-dim);margin-left:auto;">{{ r.count }} · no agent</span>
-        </div>
-        {% else %}
-        <div style="display:flex;align-items:center;gap:0.4rem;font-size:0.8rem;background:var(--bg-elevated);border:1px solid var(--border);border-radius:5px;padding:0.3rem 0.55rem;">
-          <span class="dot {% if r.status == 'running' %}dot-up{% elif r.status == 'paused' %}dot-warn{% else %}dot-down{% endif %}"></span>
-          <a href="/plugins/docker/container/{{ r.host_id }}/{{ r.name }}" style="font-weight:500;color:inherit;text-decoration:none;white-space:nowrap;overflow:hidden;text-overflow:ellipsis;" title="{{ r.name }}">{{ r.host }}</a>
-          {% if r.cpu_pct is not none %}<span style="color:var(--text-muted);font-family:ui-monospace,monospace;margin-left:auto;">{{ "%.0f" | format(r.cpu_pct) }}%</span>{% endif %}
-        </div>
-        {% endif %}
-        {% endfor %}
-      </div>
-    </div>
+{# ── Needs attention ──────────────────────────────────────────────────────── #}
+{% if attention or failing_services %}
+<section class="card problem-card" aria-labelledby="docker-problems">
+  <h2 id="docker-problems" class="section-title section-count">Needs attention <span class="tone-crit">{{ problems }}</span>
+    {%- if unhealthy %}<span class="tone-muted"> · {{ unhealthy }} unhealthy</span>{% endif %}
+    {%- if restarting %}<span class="tone-muted"> · {{ restarting }} restarting</span>{% endif %}</h2>
+  <ul class="problem-list">
+    {# A swarm service in trouble: Swarm replaces its failed tasks, so the
+       service, not any one container, is what is failing. #}
+    {% for s in failing_services %}
+    {% set e = s.last_error %}
+    <li>
+      {% if s.crash_looping %}{{ status_badge('crit', 'Crash-looping') }}
+      {% elif s.state == 'down' %}{{ status_badge('crit', 'Down') }}
+      {% else %}{{ status_badge('warn', 'Degraded') }}{% endif %}
+      <a href="/plugins/docker/service/{{ s.name }}">{{ s.name }}</a>
+      <span class="problem-detail">
+        {{- s.running }} of {{ s.desired }} running
+        {%- if s.crash_looping %} · {{ s.failures_1h }} tasks failed in the last hour{% endif %}
+        {%- if e %} · last task {{ e.state }}: {{ e.error or e.message or e.state }}{% if e.exit_code %} (exit {{ e.exit_code }}){% endif %}{% if e.node %} on {{ e.node }}{% endif %}{% endif %}
+        {%- if s.trouble_since %} · since {{ s.trouble_since|ago }}{% endif -%}
+      </span>
+      <a class="field-hint" href="/plugins/docker/service/{{ s.name }}{% if e and e.task_id %}?task={{ e.task_id }}{% endif %}#logs" style="margin-left:0.4rem;">Logs →</a>
+    </li>
     {% endfor %}
-  </div>
-</div>
+    {% for r in attention %}
+    {% set c = r.container %}
+    {% set worst = r.problems[0] %}
+    <li>
+      {{ status_badge('crit' if severity[worst] == 'critical' else 'warn', worst|replace('_', ' ')|capitalize) }}
+      <a href="/plugins/docker/container/{{ c.host_id }}/{{ c.name }}" title="{{ c.name }}">{{ r.label }}</a>
+      <span class="problem-detail">
+        {%- for p in r.problems %}{{ says[p] }}{{ " and " if not loop.last }}{% endfor %}
+        {%- if c.exit_code and worst in ('exited_nonzero', 'crash_looping', 'restarting') %} · exit {{ c.exit_code }}{% endif %}
+        {%- if r.since %} · since {{ r.since|ago }}{% endif %}
+        {%- if r.stale %} · {{ status_badge('stale') }} last report {{ c.scraped_at|ago if c.scraped_at else "never" }}{% endif -%}
+      </span>
+      <a class="field-hint" href="/plugins/docker/container/{{ c.host_id }}/{{ c.name }}/logs" style="margin-left:0.4rem;">Logs →</a>
+    </li>
+    {% endfor %}
+  </ul>
+</section>
 {% endif %}
 
-{# ── Container tables, one per host ───────────────────────────────────────── #}
-{% if host_groups %}
-{% for g in host_groups %}
-<div class="card-flush" style="margin-bottom:1.5rem;">
-  <div style="display:flex;align-items:baseline;gap:0.6rem;padding:0.5rem 0.75rem;border-bottom:1px solid var(--border);">
-    {% if g.host %}
-    <a href="/hosts/{{ g.host.id }}" style="font-weight:600;font-size:0.95rem;color:inherit;text-decoration:none;">{{ g.host.name }}</a>
-    {% else %}
-    <span style="font-weight:600;font-size:0.95rem;">{{ g.host_name }}</span>
-    {% endif %}
-    <span style="font-size:0.78rem;color:var(--text-muted);">
-      {{ g.running }} running{% if g.stopped %} · {{ g.stopped }} stopped{% endif %}
-    </span>
-  </div>
+{# ── Hosts whose Docker is not answering (agent ≥ 1.9.0) ──────────────────── #}
+{% if docker_down %}
+<section class="card problem-card" aria-labelledby="docker-unreachable">
+  <h2 id="docker-unreachable" class="section-title" style="margin-bottom:0.6rem;">Docker not answering</h2>
+  <ul class="problem-list">
+    {% for d in docker_down %}
+    <li>
+      {{ status_badge('crit', 'Error') if d.status == 'error' else status_badge('stale', 'No Docker') }}
+      <a href="/hosts/{{ d.host_id }}">{{ d.host_name }}</a>
+      <span class="problem-detail">{{ d.detail or "no reason given" }} · since {{ d.since|ago }} · its containers below are what Docker last reported</span>
+    </li>
+    {% endfor %}
+  </ul>
+</section>
+{% endif %}
+
+{# ── What failed in the last 24h: what the dashboard's "failed (24h)" counts.
+   A plain container's death, or a swarm service's failed tasks; a container
+   someone stopped (exit 137/143) is not a failure. Each links to the logs of
+   what died: a service's to its failed task's lines (#5300). ───────────── #}
+{% if failures %}
+<section id="failures" class="card" aria-labelledby="docker-failures">
+  <h2 id="docker-failures" class="section-title section-count">Failed in the last 24h <span class="tone-warn">{{ failures|length }}</span></h2>
+  <ul class="problem-list">
+    {% for x in failures %}
+    <li>
+      {{ status_badge('warn', 'Service' if x.kind == 'service' else 'Container') }}
+      {% if x.kind == 'service' %}<a href="/plugins/docker/service/{{ x.name }}">{{ x.name }}</a>
+      {% else %}<a href="/plugins/docker/container/{{ x.host_id }}/{{ x.name }}">{{ x.name }}</a>{% endif %}
+      <span class="problem-detail">
+        {{- x.count }} {{ "failure" if x.count == 1 else "failures" }}
+        {%- if x.detail %} · last: {{ x.detail }}{% endif %}
+        {%- if x.host %} · {{ x.host }}{% endif %}
+        {%- if x.at %} · {{ x.at|ago }}{% endif -%}
+      </span>
+      {% if x.kind == 'service' %}
+      <a class="field-hint" href="/plugins/docker/service/{{ x.name }}{% if x.task_id %}?task={{ x.task_id }}{% endif %}#logs" style="margin-left:0.4rem;">Logs{% if x.task_id %} of that task{% endif %} →</a>
+      {% else %}
+      <a class="field-hint" href="/plugins/docker/container/{{ x.host_id }}/{{ x.name }}/logs" style="margin-left:0.4rem;">Logs →</a>
+      {% endif %}
+    </li>
+    {% endfor %}
+  </ul>
+</section>
+{% endif %}
+
+{# ── Swarm services: one row each, problems first unless a header picked
+   another order; CPU and memory add up the running replicas (#5299). ───── #}
+{% if services %}
+<section class="card-flush" aria-labelledby="docker-services">
+  <h2 id="docker-services" class="table-caption">Swarm services <span class="tone-muted">{{ services|length }}{% if services|length != services_total %} of {{ services_total }}{% endif %}</span>
+    <a href="/plugins/docker/swarm" class="field-hint">By node →</a></h2>
   <table class="table">
     <thead>
       <tr>
-        <th>Container</th>
-        <th>Image</th>
-        <th>Ports</th>
-        <th>CPU %</th>
-        <th style="min-width:100px;">CPU history</th>
-        <th>Mem %</th>
-        <th style="min-width:100px;">Mem history</th>
+        {{ th("name", "Service") }}
+        {{ th("status", "Status", title="Problems first: a crash loop, then down, then short of replicas, then failed tasks in 24h") }}
+        {{ th("stack", "Stack", title="The part of the service's name before the first _: the stack it was deployed in") }}
+        <th title="The nodes running its replicas">Nodes</th>
+        {{ th("cpu", "CPU", "num", "Across its running replicas; Docker counts CPU per core") }}
+        {{ th("mem", "Memory", "num", "Across its running replicas") }}
+        {{ th("up", "Up", "num", "Since its newest replica started") }}
+        <th class="num" title="Tasks that failed in the last 24h; Swarm replaced them">Failed 24h</th>
       </tr>
     </thead>
     <tbody>
-      {% for sub in g.subgroups %}
-      {% if g.grouped and sub.label %}
-      <tr><td colspan="7" style="padding:0.4rem 0.75rem 0.2rem;font-size:0.72rem;color:var(--text-dim);text-transform:uppercase;letter-spacing:0.05em;font-weight:600;">{{ sub.label }}</td></tr>
-      {% endif %}
-      {% for item in sub.containers %}
-      {% set c = item.container %}
+      {% for r in services %}
       <tr>
+        <td class="cell-name"><a href="/plugins/docker/service/{{ r.name }}" title="{{ r.name }}{% if r.image %} · {{ r.image }}{% endif %}">{{ r.name }}</a></td>
         <td>
-          <div style="display:flex;align-items:center;gap:0.5rem;">
-            <span class="dot {% if c.status == 'running' %}dot-up{% elif c.status == 'paused' %}dot-warn{% else %}dot-down{% endif %}"></span>
-            <div>
-              <div style="font-weight:500;font-size:0.9rem;">
-                <a href="/plugins/docker/container/{{ c.host_id }}/{{ c.name }}" style="color:inherit;text-decoration:none;border-bottom:1px dotted var(--border-mid);">{{ c.name }}</a>
-                {% if c.health == 'healthy' %}<span title="healthy" style="color:var(--green);font-size:0.7rem;">●</span>
-                {% elif c.health == 'unhealthy' %}<span title="unhealthy" style="color:var(--red);font-size:0.7rem;">●</span>
-                {% elif c.health == 'starting' %}<span title="health: starting" style="color:var(--orange);font-size:0.7rem;">◐</span>{% endif %}
-              </div>
-              <div style="font-size:0.73rem;color:var(--text-muted);">
-                {{ c.status }}{% if item.uptime %} · up {{ item.uptime }}{% endif %}
-                {% if c.status != 'running' and c.exit_code is not none and c.exit_code != 0 %}
-                  · <span style="color:var(--red);">exit {{ c.exit_code }}{% if c.oom_killed %} (OOM){% endif %}</span>
-                {% endif %}
-                {% if c.restart_count %} · <span title="restart count" style="color:var(--orange);">⟳{{ c.restart_count }}</span>{% endif %}
-              </div>
-              {% if c.service_name or c.compose_project %}
-              <div style="font-size:0.68rem;color:var(--text-dim);margin-top:0.1rem;">
-                {{ c.service_name or c.compose_project }}
-              </div>
-              {% endif %}
-            </div>
-          </div>
+          {{ status_dot('up' if r.state == 'healthy' else ('warn' if r.state == 'degraded' else 'down')) }}
+          <span class="tone-{{ 'ok' if r.state == 'healthy' else ('warn' if r.state == 'degraded' else 'crit') }}" style="font-variant-numeric:tabular-nums;">{{ r.running }}/{{ r.desired }}</span>
+          {%- if r.crash_looping %} {{ status_badge('crit', 'Crash-looping') }}{% elif r.unhealthy %} {{ status_badge('crit', 'Unhealthy') }}{% endif %}
+          {%- if r.mode == 'global' %} <span class="tone-dim">global</span>{% endif %}
         </td>
-        <td style="font-size:0.82rem;color:var(--text-muted);max-width:200px;white-space:nowrap;overflow:hidden;text-overflow:ellipsis;">
-          {{ c.image }}
+        <td>{% if r.stack %}<a class="chip" href="{{ page_url(stack=r.stack) }}" title="Show the {{ r.stack }} stack">{{ r.stack }}</a>{% else %}<span class="tone-dim">—</span>{% endif %}</td>
+        <td class="tone-muted">
+          {{- r.nodes|join(", ") or ("" if r.ghost_nodes else "—") -}}
+          {%- if r.ghost_nodes %}{{ ", " if r.nodes }}<span title="Running on a node with no Steward agent: no figures for it">{{ r.ghost_nodes|join(", ") }} (no agent)</span>{% endif -%}
         </td>
-        <td style="font-size:0.78rem;font-family:ui-monospace,monospace;">
-          {% for p in item.ports %}
-          <div style="color:var(--text-muted);">{{ p.host_port }}:{{ p.container_port }}/{{ p.protocol }}</div>
-          {% endfor %}
+        <td class="num">{% if r.cpu_cores is not none %}{{ cores(r.cpu_cores) }}{% else %}<span class="tone-dim">—</span>{% endif %}</td>
+        <td class="num">{% if r.mem_bytes is not none %}{{ r.mem_bytes|bytes }}{% else %}<span class="tone-dim">—</span>{% endif %}</td>
+        <td class="num">{{ up(r) }}</td>
+        <td class="num">
+          {%- if r.failures_24h %}<a class="tone-warn" href="/plugins/docker/service/{{ r.name }}#tasks"
+               title="{% if r.last_error %}Last: {{ r.last_error.error or r.last_error.message or r.last_error.state }}{% if r.last_error.exit_code %} (exit {{ r.last_error.exit_code }}){% endif %}{% endif %}">{{ r.failures_24h }}</a>
+          {%- else %}<span class="tone-dim">—</span>{% endif -%}
         </td>
-        <td style="font-family:ui-monospace,monospace;font-size:0.88rem;">
-          {% if c.cpu_pct is not none %}
-          <span style="color:{% if c.cpu_pct > 80 %}var(--red){% elif c.cpu_pct > 50 %}var(--orange){% else %}var(--text){% endif %};">
-            {{ "%.1f" | format(c.cpu_pct) }}%
-          </span>
-          {% else %}
-          <span style="color:var(--text-dim);">—</span>
-          {% endif %}
-        </td>
-        <td>{{ item.sparkline_cpu | safe }}</td>
-        <td style="font-family:ui-monospace,monospace;font-size:0.88rem;">
-          {% if c.mem_pct is not none %}
-          <span style="color:{% if c.mem_pct > 90 %}var(--red){% elif c.mem_pct > 70 %}var(--orange){% else %}var(--text){% endif %};">
-            {{ "%.1f" | format(c.mem_pct) }}%
-          </span>
-          {% else %}
-          <span style="color:var(--text-dim);">—</span>
-          {% endif %}
-        </td>
-        <td>{{ item.sparkline_mem | safe }}</td>
       </tr>
       {% endfor %}
-      {% endfor %}
     </tbody>
   </table>
-</div>
-{% endfor %}
-{% elif not swarm_services %}
-<div class="card" style="text-align:center;padding:3rem;">
-  <div style="color:var(--text-muted);font-size:0.9rem;">
+</section>
+{% endif %}
+
+{# ── Containers outside any swarm service, every host in one table (#5301);
+   a compose project is its stack. ─────────────────────────────────────── #}
+{% if containers %}
+<section class="card-flush" aria-labelledby="docker-containers">
+  <h2 id="docker-containers" class="table-caption">{{ "Other containers" if services_total else "Containers" }} <span class="tone-muted">{{ containers|length }}{% if containers|length != containers_total %} of {{ containers_total }}{% endif %}</span></h2>
+  <table class="table">
+    <thead>
+      <tr>
+        {{ th("name", "Container") }}
+        {{ th("status", "Status", title="Problems first: unhealthy or restarting, then died with a failing exit, then stale; then running, then stopped") }}
+        <th>Host</th>
+        {{ th("stack", "Stack", title="Its compose project") }}
+        <th>Image</th>
+        {{ th("cpu", "CPU", "num", "Docker counts CPU per core: 1 core = one core fully busy") }}
+        {{ th("mem", "Memory", "num", "Used, and the limit when one is set") }}
+        {{ th("up", "Up", "num", "Since it started") }}
+      </tr>
+    </thead>
+    <tbody>
+      {% for r in containers %}
+      <tr>
+        <td class="cell-name">
+          <a href="/plugins/docker/container/{{ r.host_id }}/{{ r.container_name }}" title="{{ r.container_name }}">{{ r.name }}</a>
+          {%- if r.ports %}<div class="cell-sub mono">{% for p in r.ports %}{{ p.host_port }}:{{ p.container_port }}/{{ p.protocol }}{{ " " if not loop.last }}{% endfor %}</div>{% endif %}
+        </td>
+        <td>
+          {{ status_dot('stale' if r.stale else ('up' if r.running else ('pending' if r.status == 'paused' else 'down'))) }}
+          {%- if r.health == 'unhealthy' %} {{ status_badge('crit', 'Unhealthy') }}
+          {%- elif r.health == 'starting' %} {{ status_badge('pending', 'Starting') }}
+          {%- else %} <span class="tone-muted">{{ r.status }}</span>{% endif %}
+          {%- if not r.running and r.exit_code is not none and r.exit_code != 0 %}
+          <span class="tone-{{ 'crit' if r.exit_failed else 'dim' }}" title="{{ 'last exit code' if r.exit_failed else 'stopped by a signal; not a failure' }}">exit {{ r.exit_code }}{% if r.oom_killed %} (OOM){% endif %}</span>
+          {%- endif %}
+          {%- if r.restart_count %} <span class="tone-warn" title="Restarted {{ r.restart_count }} times">⟳{{ r.restart_count }}</span>{% endif %}
+          {%- if r.stale %} {{ status_badge('stale', title="Last report " ~ (r.scraped_at.strftime('%Y-%m-%d %H:%M UTC') if r.scraped_at else 'never') ~ "; what is shown may no longer be true") }}{% endif %}
+        </td>
+        <td class="tone-muted">{{ r.host }}</td>
+        <td>{% if r.stack %}<a class="chip" href="{{ page_url(stack=r.stack) }}" title="Show the {{ r.stack }} stack">{{ r.stack }}</a>{% else %}<span class="tone-dim">—</span>{% endif %}</td>
+        <td class="cell-image" title="{{ r.image }}">{{ r.image }}</td>
+        <td class="num">{% if r.running %}{{ cpu(r.usage) }}{% else %}<span class="tone-dim">—</span>{% endif %}</td>
+        <td class="num">{% if r.running %}{{ mem(r.usage) }}{% else %}<span class="tone-dim">—</span>{% endif %}</td>
+        <td class="num">{{ up(r) }}</td>
+      </tr>
+      {% endfor %}
+    </tbody>
+  </table>
+</section>
+{% endif %}
+
+{% if not services and not containers %}
+<div class="card">
+  {% if services_total or containers_total %}
+  <p class="tone-muted" style="margin:0;">Nothing matches these filters. <a href="/plugins/docker/">Show everything</a></p>
+  {% else %}
+  <p class="tone-muted" style="margin:0;">
     No containers reported yet. Containers are collected by the Steward host agent —
     deploy the agent to a host running Docker (Hosts → the host → agent panel) and
-    its containers will appear here under that host.
-  </div>
+    its containers will appear here.
+  </p>
+  {% endif %}
 </div>
 {% endif %}

plugins/docker/templates/docker/service.html

@@ -0,0 +1,160 @@
+{# docker/service.html — one swarm service (#5302, #5310): a header that says
+   whether it is well and what it runs; its last failure only while it is in
+   trouble; its CPU and memory over time; one table of its tasks, the running
+   ones first with what they use; and the logs of every task it has run.
+   A failure's "Logs →" lands on #logs with ?task= picked. #}
+{% extends "base.html" %}
+{% from "_macros.html" import crumbs, status_badge %}
+{% from "docker/_usage.html" import cpu, mem %}
+{% block title %}{{ name }} — Docker — Steward{% endblock %}
+{% block breadcrumb %}{{ crumbs([("Docker", "/plugins/docker/"), (name, "")]) }}{% endblock %}
+{% block content %}
+
+{% if service is none %}
+<div class="card empty-state">
+  <strong>{{ "No longer running" if task_choices else "Service not found" }}</strong>
+  {% if task_choices %}
+  <span>No swarm service named <code>{{ name }}</code> is reported now; it may have been removed.
+    The logs Steward still keeps from its tasks are below.</span>
+  {% else %}
+  <span>No swarm service named <code>{{ name }}</code> is reported, and Steward holds no logs or
+    events for one. It may have been removed. <a href="/plugins/docker/">Back to Docker</a></span>
+  {% endif %}
+</div>
+{% else %}
+
+{# ── Header: the verdict, then what it is (#5310). Each fact has one home:
+   what it uses is in Resource history, its tasks and failures in Tasks. #}
+{% set troubled = row.crash_looping or row.state != 'healthy' %}
+{% set tone = 'ok' if row.state == 'healthy' else ('warn' if row.state == 'degraded' else 'crit') %}
+{% set image = (svc.image if svc and svc.image else row.image) %}
+{% set e = row.last_error %}
+<div class="page-header">
+  <div style="min-width:0;">
+    <h1 class="page-title" style="overflow-wrap:anywhere;">{{ name }}</h1>
+    <p class="page-subtitle">
+      {% if row.crash_looping %}{{ status_badge('crit', 'Crash-looping') }}
+      {% elif row.state == 'down' %}{{ status_badge('crit', 'Down') }}
+      {% elif row.state == 'degraded' %}{{ status_badge('warn', 'Degraded') }}
+      {% else %}{{ status_badge('ok', 'Healthy') }}{% endif %}
+      <span class="tone-{{ tone }}">{{ row.running }} of {{ row.desired }} running</span>
+      {%- if troubled and row.trouble_since %} · <span class="tone-{{ tone }}">in trouble since {{ row.trouble_since|ago }}</span>{% endif %}
+      · {{ row.mode }}
+      {%- if stack %} · stack <a class="chip" href="/plugins/docker/?stack={{ stack }}">{{ stack }}</a>{% endif %}
+      {%- if image %} · <span class="mono" style="overflow-wrap:anywhere;" title="{{ image }}">{{ image.split("@")[0] }}</span>{% endif %}
+    </p>
+    {% if (not troubled and e) or svc %}
+    <p class="page-subtitle tone-dim" style="margin-top:0.2rem;">
+      {%- if not troubled and e %}Last failed {% if e.age_s is not none %}{{ e.age_s|duration }} ago{% endif %}
+      {%- if e.exit_code %} (exit {{ e.exit_code }}){% endif %}
+      {%- if e.task_id %} · <a href="?task={{ e.task_id }}#logs">its logs</a>{% endif %} · {% endif -%}
+      {%- if svc %}reported {{ svc.updated_at|ago }} by the manager's agent{% endif -%}
+    </p>
+    {% endif %}
+  </div>
+  <div class="page-actions"><a class="btn btn-sm btn-ghost" href="#logs">Logs</a></div>
+</div>
+
+{# ── What is wrong, only while it is (#5310) ──────────────────────────────── #}
+{% if troubled and e %}
+<section class="card problem-card" aria-labelledby="svc-last-error">
+  <h2 id="svc-last-error" class="section-title" style="margin-bottom:0.4rem;">Last failed task</h2>
+  <p style="margin:0;">
+    {{ e.error or e.message or e.state }}{% if e.exit_code %} <span class="tone-crit">(exit {{ e.exit_code }})</span>{% endif %}
+    {%- if e.node %} · on {{ e.node }}{% endif %}{% if e.age_s is not none %} · {{ e.age_s|duration }} ago{% endif %}
+    {%- if e.task_id %} · <a href="?task={{ e.task_id }}#logs">its logs →</a>{% endif %}
+  </p>
+</section>
+{% endif %}
+
+{# ── Resource history, summed across replicas ────────────────────────────── #}
+<section class="card" aria-labelledby="svc-history">
+  <div style="display:flex;align-items:center;justify-content:space-between;gap:1rem;margin-bottom:0.75rem;flex-wrap:wrap;">
+    <h2 id="svc-history" class="section-title" style="margin:0;">Resource history <span class="tone-dim">across running replicas</span></h2>
+    {% include "_time_range.html" %}
+  </div>
+  <div hx-get="/plugins/docker/service/{{ name }}/history"
+       hx-trigger="load, rangeChange from:body"
+       hx-include="#time-range"
+       hx-swap="innerHTML">
+    <p class="field-hint">Loading…</p>
+  </div>
+</section>
+
+{# ── Tasks: the running ones with what they use, then Swarm's record ─────── #}
+<section id="tasks" class="card-flush" aria-labelledby="svc-tasks">
+  <h2 id="svc-tasks" class="table-caption">Tasks
+    <span class="tone-muted">running first, then Swarm's most recent</span>
+    {%- if row.failures_24h %} <span class="tone-warn">· {{ row.failures_24h }} failed in 24h</span>{% endif %}</h2>
+  {% if tasks %}
+  <table class="table">
+    <thead><tr>
+      <th title="The replica slot this task filled">Slot</th><th>Task</th><th>Node</th><th>State</th>
+      <th class="num" title="Docker counts CPU per core">CPU</th><th class="num">Memory</th>
+      <th class="num">Exit</th><th>Why</th><th class="num" title="When it started, for a running task; else when it reached its state">When</th>
+    </tr></thead>
+    <tbody>
+      {% for t in tasks %}
+      <tr>
+        <td>{{ t.slot or "—" }}</td>
+        <td class="mono" title="{{ t.container or t.task_id }}">
+          {%- if t.task_id %}<a href="?task={{ t.task_id }}#logs" title="Its logs">{{ t.task_id[:10] }}</a>
+          {%- if t.container and t.host_id %} <a class="tone-dim" href="/plugins/docker/container/{{ t.host_id }}/{{ t.container }}" title="The container's own page">↗</a>{% endif %}
+          {%- elif t.container and t.host_id %}<a href="/plugins/docker/container/{{ t.host_id }}/{{ t.container }}">{{ t.container|truncate(24, true, "…") }}</a>
+          {%- else %}<span class="tone-dim">—</span>{% endif -%}
+        </td>
+        <td>{% if t.host_id %}<a href="/hosts/{{ t.host_id }}">{{ t.node }}</a>{% else %}{{ t.node }}{% endif %}</td>
+        <td>
+          {%- if t.failed %}{{ status_badge('crit', t.state|capitalize) }}
+          {%- elif t.current %}{{ status_badge('ok', 'Running') }}
+          {%- else %}<span class="tone-muted">{{ t.state }}</span>{% endif %}
+          {%- if t.health == 'unhealthy' %} {{ status_badge('crit', 'Unhealthy') }}{% elif t.health == 'starting' %} {{ status_badge('pending', 'Starting') }}{% endif %}
+          {%- if t.desired and t.desired != t.state and not t.current %} <span class="tone-dim" title="What Swarm wants of it">→ {{ t.desired }}</span>{% endif -%}
+        </td>
+        <td class="num">{% if t.usage %}{{ cpu(t.usage) }}{% else %}<span class="tone-dim">—</span>{% endif %}</td>
+        <td class="num">{% if t.usage %}{{ mem(t.usage) }}{% else %}<span class="tone-dim">—</span>{% endif %}</td>
+        <td class="num">{% if t.exit_code %}<span class="tone-{{ 'crit' if t.failed else 'dim' }}">{{ t.exit_code }}</span>{% else %}<span class="tone-dim">—</span>{% endif %}</td>
+        <td style="overflow-wrap:anywhere;">{{ t.why or "" }}</td>
+        <td class="num">{% if t.at %}{{ t.at|ago }}{% else %}<span class="tone-dim">—</span>{% endif %}</td>
+      </tr>
+      {% endfor %}
+    </tbody>
+  </table>
+  {% else %}
+  <p class="tone-muted" style="margin:0;padding:1rem;">No task is running, and Swarm's record of its tasks has not been reported yet (the manager's agent reports it from version 1.8.0).</p>
+  {% endif %}
+</section>
+
+{% endif %}
+
+{# ── Logs across every task, newest first ─────────────────────────────────
+   Shown for a removed service too, while its lines are still kept. #}
+{% if service is not none or task_choices %}
+<section id="logs" class="card-flush" aria-labelledby="svc-logs">
+  <h2 id="svc-logs" class="table-caption">Logs <span class="tone-muted">every task, newest first</span></h2>
+  <form id="svc-log-controls" class="filter-bar" onsubmit="return false;" style="padding:0.75rem 1rem 0;">
+    <input type="search" name="q" class="input" placeholder="Filter lines…" autocomplete="off" aria-label="Filter log lines">
+    {% if task_choices|length > 1 %}
+    <select name="task" aria-label="Task">
+      <option value="">All tasks</option>
+      {% for tid, label in task_choices %}<option value="{{ tid }}"{% if tid == focus_task %} selected{% endif %}>{{ label }}</option>{% endfor %}
+    </select>
+    {% elif focus_task %}<input type="hidden" name="task" value="{{ focus_task }}">{% endif %}
+    <select name="stream" aria-label="Stream">
+      <option value="all">stdout and stderr</option>
+      <option value="stdout">stdout</option>
+      <option value="stderr">stderr</option>
+    </select>
+  </form>
+  <div id="svc-log-lines"
+       hx-get="/plugins/docker/service/{{ name }}/logs/lines"
+       hx-trigger="load, every 5s, change from:#svc-log-controls, keyup changed delay:400ms from:#svc-log-controls"
+       hx-include="#svc-log-controls"
+       hx-swap="innerHTML"
+       role="log" aria-live="polite" tabindex="0"
+       style="max-height:60vh;overflow:auto;padding:0.5rem 1rem 0.75rem;font-family:var(--fs-font-mono);font-size:0.8rem;line-height:1.5;">
+    <p class="field-hint">Loading…</p>
+  </div>
+</section>
+{% endif %}
+{% endblock %}

plugins/docker/templates/docker/swarm.html

@@ -1,87 +1,103 @@
-{# docker/swarm.html — Swarm topology: services, replica health, nodes, placement #}
+{# docker/swarm.html — the Swarm laid out by node (#5303). A verdict, the
+   services short of replicas, then one card per node: a node that is down or
+   not taking work first, then managers. Each card says what its containers use
+   against what the machine has, and lists the services placed on it. #}
 {% extends "base.html" %}
-{% from "_macros.html" import crumbs %}
+{% from "_macros.html" import crumbs, status_badge, status_dot %}
+{% from "docker/_usage.html" import cores as fmt_cores %}
 {% block title %}Swarm — Docker — Steward{% endblock %}
 {% block breadcrumb %}{{ crumbs([("Docker", "/plugins/docker/"), ("Swarm", "")]) }}{% endblock %}
 {% block content %}
 
-<h1 class="page-title" style="margin-bottom:1.5rem;">Swarm</h1>
+<h1 class="page-title">Swarm</h1>
 
 {% if swarms %}
 {% for g in swarms %}
-<div style="margin-bottom:2rem;">
-  <div style="display:flex;align-items:baseline;gap:0.6rem;margin-bottom:0.75rem;flex-wrap:wrap;">
-    <span style="font-weight:600;font-size:0.95rem;">Swarm</span>
-    <span style="font-size:0.78rem;color:var(--text-muted);">
-      reported by {{ g.managers | length }} manager{{ 's' if g.managers | length != 1 }} · {{ g.managers | join(", ") }}
-    </span>
-  </div>
+{% set down_nodes = g.nodes|rejectattr("ok")|list %}
+<section class="swarm-group" aria-label="Swarm reported by {{ g.managers|join(', ') }}">
+  <p class="page-verdict">
+    {% if down_nodes or g.troubled %}<strong class="tone-crit">{{ down_nodes|length + g.troubled|length }} need{{ "s" if (down_nodes|length + g.troubled|length) == 1 }} attention</strong>
+    {% else %}<strong class="tone-ok">Every node is up and taking work</strong>{% endif %}
+    · {{ g.nodes|length }} node{{ "s" if g.nodes|length != 1 }}
+    · <span class="tone-{{ 'ok' if g.services_healthy == g.services_total else 'warn' }}">{{ g.services_healthy }}/{{ g.services_total }}</span> services healthy
+    <span class="tone-dim">· reported by {{ g.managers|join(", ") }}</span>
+  </p>
 
-  {# ── Services ─────────────────────────────────────────────────────────── #}
-  <div class="card-flush" style="margin-bottom:1rem;">
-    <table class="table">
-      <thead>
-        <tr>
-          <th>Service</th><th>Mode</th><th>Replicas</th><th>Image</th><th>Placement</th>
-        </tr>
-      </thead>
-      <tbody>
-        {% for s in g.services %}
-        <tr>
-          <td style="font-weight:500;font-size:0.9rem;">{{ s.name }}</td>
-          <td style="font-size:0.82rem;color:var(--text-muted);">{{ s.mode }}</td>
-          <td style="font-family:ui-monospace,monospace;font-size:0.88rem;">
-            <span style="color:{% if s.healthy %}var(--green){% elif s.down %}var(--red){% elif s.degraded %}var(--orange){% else %}var(--text-muted){% endif %};">
-              {{ s.running }}/{{ s.desired }}
-            </span>
-          </td>
-          <td style="font-size:0.8rem;color:var(--text-muted);max-width:220px;white-space:nowrap;overflow:hidden;text-overflow:ellipsis;">{{ s.image or "—" }}</td>
-          <td style="font-size:0.78rem;color:var(--text-muted);">
-            {% for p in s.placement %}{{ p.node }} ({{ p.running }}){% if not loop.last %}, {% endif %}{% else %}—{% endfor %}
-          </td>
-        </tr>
-        {% else %}
-        <tr><td colspan="5" style="color:var(--text-muted);font-size:0.85rem;text-align:center;padding:1.5rem;">No services in this swarm.</td></tr>
-        {% endfor %}
-      </tbody>
-    </table>
-  </div>
+  {% if down_nodes or g.troubled %}
+  <section class="card problem-card" aria-label="Needs attention">
+    <ul class="problem-list">
+      {% for n in down_nodes %}
+      <li>
+        {{ status_badge('crit', 'Node ' ~ n.status) if n.status != 'ready' else status_badge('warn', n.availability|capitalize) }}
+        <span>{{ n.hostname }}</span>
+        <span class="problem-detail">{{ n.role }} · {{ n.task_count }} task{{ "s" if n.task_count != 1 }} placed on it</span>
+      </li>
+      {% endfor %}
+      {% for s in g.troubled %}
+      <li>
+        {% if s.crash_looping %}{{ status_badge('crit', 'Crash-looping') }}{% elif s.state == 'down' %}{{ status_badge('crit', 'Down') }}{% else %}{{ status_badge('warn', 'Degraded') }}{% endif %}
+        <a href="/plugins/docker/service/{{ s.name }}">{{ s.name }}</a>
+        {% set e = s.last_error %}
+        <span class="problem-detail">
+          {{- s.running }} of {{ s.desired }} running
+          {%- if e %} · last task {{ e.state }}: {{ e.error or e.message or e.state }}{% if e.node %} on {{ e.node }}{% endif %}{% endif %}
+          {%- if s.trouble_since %} · since {{ s.trouble_since|ago }}{% endif -%}
+        </span>
+      </li>
+      {% endfor %}
+    </ul>
+  </section>
+  {% endif %}
 
-  {# ── Nodes ────────────────────────────────────────────────────────────── #}
-  <div class="card-flush">
-    <table class="table">
-      <thead>
-        <tr><th>Node</th><th>Role</th><th>Availability</th><th>Status</th></tr>
-      </thead>
-      <tbody>
-        {% for n in g.nodes %}
-        <tr>
-          <td style="font-weight:500;font-size:0.9rem;">
-            {{ n.hostname or n.node_id }}
-            {% if n.leader %}<span title="cluster leader" style="font-size:0.68rem;color:var(--accent);border:1px solid var(--accent);border-radius:3px;padding:0 0.3rem;margin-left:0.3rem;">leader</span>{% endif %}
-          </td>
-          <td style="font-size:0.82rem;color:var(--text-muted);">{{ n.role }}</td>
-          <td style="font-size:0.82rem;color:{% if n.availability == 'active' %}var(--text){% else %}var(--orange){% endif %};">{{ n.availability }}</td>
-          <td style="font-size:0.82rem;">
-            <span class="dot {% if n.status == 'ready' %}dot-up{% else %}dot-down{% endif %}"></span>
-            <span style="color:{% if n.status == 'ready' %}var(--green){% else %}var(--red){% endif %};">{{ n.status }}</span>
-          </td>
-        </tr>
+  <div class="node-grid">
+    {% for n in g.nodes %}
+    <section class="card node-card{% if not n.ok %} problem-card{% endif %}" aria-labelledby="node-{{ n.node_id }}">
+      <h2 id="node-{{ n.node_id }}" class="node-card-head">
+        {{ status_dot('up' if n.status == 'ready' else 'down') }}
+        {% if n.host_id %}<a href="/hosts/{{ n.host_id }}">{{ n.hostname }}</a>{% else %}{{ n.hostname }}{% endif %}
+        <span class="chip">{{ "leader" if n.leader else n.role }}</span>
+        {% if n.availability != 'active' %}{{ status_badge('warn', n.availability|capitalize, title="Swarm places no new tasks here") }}{% endif %}
+      </h2>
+      <p class="node-card-figures">
+        {% if n.cores_used is not none %}
+        Containers use <strong>{{ fmt_cores(n.cores_used) }}</strong>{% if n.cores %} of {{ n.cores }}{% endif %}
+        · <strong>{{ n.mem_used|bytes }}</strong>{% if n.mem %} of {{ n.mem|bytes }}{% endif %}
+        {% elif n.agent %}
+        <span class="tone-muted">No swarm task reports figures here yet.</span>
         {% else %}
-        <tr><td colspan="4" style="color:var(--text-muted);font-size:0.85rem;text-align:center;padding:1.5rem;">No nodes reported.</td></tr>
-        {% endfor %}
-      </tbody>
-    </table>
+        <span class="tone-muted" title="Install the Steward agent on this node to see what its tasks use">No Steward agent: what runs here is Swarm's placement, without figures.</span>
+        {% endif %}
+      </p>
+      {% if n.tasks %}
+      <table class="table node-tasks">
+        <thead><tr><th>Service</th><th class="num">Tasks</th><th class="num">CPU</th><th class="num">Memory</th></tr></thead>
+        <tbody>
+          {% for t in n.tasks %}
+          <tr>
+            <td class="cell-name">
+              {%- if t.state != 'healthy' %}{{ status_dot('down' if t.state == 'down' else 'warn') }} {% endif -%}
+              <a href="/plugins/docker/service/{{ t.name }}" title="{{ t.name }}">{{ t.name }}</a>
+            </td>
+            <td class="num">{{ t.tasks }}</td>
+            <td class="num">{% if t.cores is not none %}{{ fmt_cores(t.cores) }}{% else %}<span class="tone-dim">—</span>{% endif %}</td>
+            <td class="num">{% if t.mem_bytes is not none %}{{ t.mem_bytes|bytes }}{% else %}<span class="tone-dim">—</span>{% endif %}</td>
+          </tr>
+          {% endfor %}
+        </tbody>
+      </table>
+      {% else %}
+      <p class="tone-muted" style="margin:0;">No tasks placed here.</p>
+      {% endif %}
+    </section>
+    {% endfor %}
   </div>
-</div>
+</section>
 {% endfor %}
 {% else %}
-<div class="card" style="text-align:center;padding:3rem;">
-  <div style="color:var(--text-muted);font-size:0.9rem;">
-    No Swarm topology reported. A Steward host agent running on a Swarm
-    <strong>manager</strong> reports services, nodes, and task placement here —
-    workers and non-swarm hosts contribute only their local containers.
-  </div>
+<div class="card empty-state">
+  No Swarm topology reported. A Steward host agent running on a Swarm
+  <strong>manager</strong> reports services, nodes and task placement here;
+  workers and hosts outside a swarm contribute only their own containers.
 </div>
 {% endif %}
 {% endblock %}

plugins/docker/templates/docker/widget.html

@@ -1,67 +1,80 @@
-{# docker/widget.html — dashboard widget: container status overview, by host #}
+{# docker/widget.html — the dashboard Docker widget (#5252), in the widget-shape
+   order: a verdict line; the services and containers in trouble; what failed
+   in the last 24h, by name; then everything healthy, one line each, packed
+   into as many columns as the widget is wide. Resource use is the Busiest
+   Containers widget's, so it isn't repeated here. #}
+{% from "_macros.html" import status_dot %}
 {% if total_count == 0 %}
-<p class="empty" style="padding:0.5rem 0;font-size:0.85rem;">No containers reported yet.</p>
+<div class="w-verdict" style="color:var(--text-muted);">No containers reported yet.</div>
 {% else %}
-
-<div style="display:flex;gap:1.25rem;margin-bottom:0.65rem;">
-  <div>
-    <span style="font-size:1.4rem;font-weight:700;line-height:1;color:var(--green);">{{ running_count }}</span>
-    <span style="font-size:0.78rem;color:var(--text-muted);margin-left:0.25rem;">running</span>
-  </div>
-  <div title="Distinct containers with a die or OOM event in the last 24h">
-    <span style="font-size:1.4rem;font-weight:700;line-height:1;color:{% if failed_24h %}var(--red){% else %}var(--text-muted){% endif %};">{{ failed_24h }}</span>
-    <span style="font-size:0.78rem;color:var(--text-muted);margin-left:0.25rem;">failed (24h)</span>
-  </div>
+{% set wrong = bad_services|length + bad_containers|length %}
+<div class="w-verdict {{ 'tone-crit' if wrong else ('tone-warn' if failures else 'tone-ok') }}">
+  {%- if services_total %}{{ services_healthy }}/{{ services_total }} services healthy · {% endif -%}
+  {{ running_count }} other container{{ 's' if running_count != 1 }} running ·
+  {% if failures %}<a href="/plugins/docker/#failures" style="color:inherit;">{{ failures|length }} failed in 24h</a>{% else %}nothing failed in 24h{% endif -%}
 </div>
 
-{# Swarm services — collapsed, each with its replicas' host chips (dashed = a
-   node with no Steward agent, count-only from placement). #}
-{% if swarm_services %}
-<div style="font-size:0.72rem;color:var(--text-dim);text-transform:uppercase;letter-spacing:0.05em;margin:0.5rem 0 0.2rem;">Swarm services</div>
-<div style="display:grid;gap:0.4rem;margin-bottom:0.4rem;">
-  {% for s in swarm_services %}
-  <div>
-    <div style="display:flex;align-items:center;gap:0.4rem;font-size:0.82rem;">
-      <span class="dot {% if s.state == 'healthy' %}dot-up{% elif s.state == 'degraded' %}dot-warn{% else %}dot-down{% endif %}"></span>
-      <span style="font-weight:500;min-width:0;white-space:nowrap;overflow:hidden;text-overflow:ellipsis;">{{ s.name }}</span>
-      <span style="color:var(--text-muted);font-family:ui-monospace,monospace;font-size:0.74rem;margin-left:auto;flex-shrink:0;">{{ s.running }}/{{ s.desired }}</span>
-    </div>
-    <div style="display:flex;flex-wrap:wrap;gap:0.25rem;margin:0.2rem 0 0 1rem;">
-      {% for r in s.replicas %}
-      {% if r.ghost %}
-      <span title="no Steward agent on this node — count from swarm placement" style="font-size:0.68rem;color:var(--text-dim);background:var(--bg-elevated);border:1px dashed var(--border-mid);border-radius:4px;padding:0.05rem 0.35rem;">{{ r.host }} ×{{ r.count }}</span>
-      {% else %}
-      <span style="font-size:0.68rem;color:var(--text-muted);background:var(--bg-elevated);border:1px solid var(--border);border-radius:4px;padding:0.05rem 0.35rem;">{{ r.host }}</span>
-      {% endif %}
-      {% endfor %}
-    </div>
-  </div>
-  {% endfor %}
-</div>
-{% endif %}
-
-{% for g in host_groups %}
-{% if multi_host %}
-<div style="font-size:0.72rem;color:var(--text-dim);text-transform:uppercase;letter-spacing:0.05em;margin:0.5rem 0 0.2rem;">{{ g.host_name }}</div>
-{% endif %}
-<div style="display:grid;gap:2px;">
-  {% for c in g.containers %}
-  <div style="display:flex;align-items:center;gap:0.5rem;font-size:0.82rem;padding:0.15rem 0;">
-    <span class="dot {% if c.status == 'running' %}dot-up{% elif c.status == 'paused' %}dot-warn{% else %}dot-down{% endif %}"></span>
-    <a href="/plugins/docker/container/{{ c.host_id }}/{{ c.name }}" style="flex:1;min-width:0;white-space:nowrap;overflow:hidden;text-overflow:ellipsis;color:inherit;text-decoration:none;">
-      {{ c.name }}{% if c.health == 'unhealthy' %}<span title="unhealthy" style="color:var(--red);"> ●</span>{% elif c.health == 'healthy' %}<span title="healthy" style="color:var(--green);"> ●</span>{% endif %}{% if c.restart_count %}<span title="restarts" style="color:var(--orange);font-size:0.7rem;"> ⟳{{ c.restart_count }}</span>{% endif %}
-    </a>
-    {% if c.status != 'running' and c.exit_code is not none and c.exit_code != 0 %}
-    <span title="last exit code" style="font-size:0.7rem;color:var(--red);flex-shrink:0;">exit {{ c.exit_code }}</span>
-    {% endif %}
-    {% if c.cpu_pct is not none %}
-    <span style="font-size:0.72rem;color:var(--text-muted);font-family:ui-monospace,monospace;flex-shrink:0;">
-      {{ "%.1f" | format(c.cpu_pct) }}%
+{% if wrong %}
+<div class="w-section">Needs a look</div>
+{% for s in bad_services %}
+<div class="w-item">
+  {{ status_dot('degraded' if s.state == 'degraded' else 'down') }}
+  <span class="w-label" title="{{ s.name }}">{{ s.name }}
+    <span class="w-sub {{ 'tone-crit' if s.state != 'degraded' else 'tone-warn' }}">
+      {%- if s.crash_looping %}crash-looping: {{ s.failures_1h }} failed in the last hour{% else %}{{ s.running }} of {{ s.desired }} running{% endif %}
+      {%- if s.trouble_since %}, since {{ s.trouble_since|ago }}{% endif %}
+      {%- if s.last_error and (s.last_error.error or s.last_error.message) %} · {{ s.last_error.error or s.last_error.message }}{% endif -%}
     </span>
-    {% endif %}
-  </div>
-  {% endfor %}
+  </span>
+  <span class="w-meta" title="replicas running / wanted">{{ s.running }}/{{ s.desired }}</span>
+</div>
+{% endfor %}
+{% for c in bad_containers %}
+<div class="w-item">
+  {{ status_dot('warn' if c.status == 'restarting' else 'down') }}
+  <span class="w-label" title="{{ c.name }}">
+    {%- if session.user_id %}<a href="/plugins/docker/container/{{ c.host_id }}/{{ c.name }}">{{ c.name }}</a>{% else %}{{ c.name }}{% endif %}
+    <span class="w-sub tone-crit">
+      {%- if c.health == 'unhealthy' %}unhealthy{% elif c.status == 'restarting' %}restarting{% else %}exited {{ c.exit_code }}{% if c.oom_killed %}, out of memory{% endif %}{% endif %} · {{ host_name.get(c.host_id, c.host_id) }}
+    </span>
+  </span>
 </div>
 {% endfor %}
+{% endif %}
 
+{% if failures %}
+<div class="w-section">Failed in the last 24h</div>
+{% for f in failures[:5] %}
+<div class="w-item">
+  {{ status_dot('failed') }}
+  <span class="w-label" title="{{ f.name }}">{{ f.name }}
+    <span class="w-sub">{% if f.count > 1 %}{{ f.count }} times, last {% endif %}{{ f.at|ago }}{% if f.detail %} · {{ f.detail }}{% endif %} · {{ f.host }}</span>
+  </span>
+</div>
+{% endfor %}
+{% if failures|length > 5 %}<div class="w-more">+{{ failures|length - 5 }} more · <a href="/plugins/docker/#failures">the full list</a></div>{% endif %}
+{% endif %}
+
+{% if healthy %}
+<div class="w-section">Healthy</div>
+<div class="w-grid">
+  {% for h in healthy %}
+  <div class="w-item">
+    {% if h.kind == 'service' %}{% set s = h.s %}
+    {{ status_dot('up') }}
+    <span class="w-label" title="{{ s.name }} — on {{ s.replicas|map(attribute='host')|unique|join(', ') }}">{{ s.name }}
+      <span style="color:var(--text-dim);"> · {{ s.replicas|map(attribute='host')|unique|join(', ') }}</span></span>
+    <span class="w-meta" title="replicas running / wanted">{{ s.running }}/{{ s.desired }}</span>
+    {% else %}{% set c = h.c %}
+    {{ status_dot('up' if h.kind == 'container' else 'unknown') }}
+    <span class="w-label" title="{{ c.name }}">
+      {%- if session.user_id %}<a href="/plugins/docker/container/{{ c.host_id }}/{{ c.name }}">{{ c.name }}</a>{% else %}{{ c.name }}{% endif %}
+      <span style="color:var(--text-dim);"> · {{ host_name.get(c.host_id, c.host_id) }}</span></span>
+    {% if h.kind == 'stopped' %}<span class="w-meta">stopped</span>{% endif %}
+    {% endif %}
+  </div>
+  {% endfor %}
+</div>
+{% if healthy_more %}<div class="w-more">+{{ healthy_more }} more healthy · <a href="/plugins/docker/">all of Docker</a></div>{% endif %}
+{% endif %}
 {% endif %}

plugins/docker/templates/docker/widget_resources.html

@@ -1,29 +1,51 @@
-{# docker/widget_resources.html — dashboard widget: the busiest running containers
-   by CPU, each with labeled CPU + memory utilisation bars. #}
+{# docker/widget_resources.html — the dashboard Busiest Containers widget
+   (#5253, #5214): a verdict naming the busiest, then the top containers by CPU
+   (averaged over the last few minutes, so the order holds still). Each figure
+   carries its unit, beside bars of the last few hours. A row's bars are scaled
+   to its own peak, but never below a floor (1 core, 1 GiB by default), so a
+   container idling at 0.05 cores reads flat and a spike means real load. #}
+{% macro cores(v) -%}{{ "%.2f"|format(v) if v < 1 else "%.1f"|format(v) }}{%- endmacro %}
+{% macro bars(b, peak_text, roof_text) -%}
+<span class="w-bars w-wide-only" aria-hidden="true"
+      title="{% if b.peak is none %}No samples in the last {{ hours }}h{% else %}Peak {{ peak_text }} in the last {{ hours }}h; bars reach the top at {{ roof_text }}{% endif %}">
+  {%- for h in b.bars %}<i{% if h is none %} class="gap"{% else %} style="height:{{ [h, 4]|max }}%"{% endif %}></i>{% endfor -%}
+</span>
+{%- endmacro %}
 {% if not rows %}
-<div style="color:var(--text-muted);font-size:0.85rem;padding:0.5rem 0;">No running containers.</div>
+<div class="w-verdict" style="color:var(--text-muted);">No running containers.</div>
 {% else %}
-{% macro bar(label, pct, warn, crit) %}
-<div style="display:flex;align-items:center;gap:0.45rem;margin-bottom:2px;">
-  <span style="font-size:0.62rem;color:var(--text-dim);width:1.9rem;flex-shrink:0;letter-spacing:0.03em;">{{ label }}</span>
-  <div style="flex:1;height:6px;background:var(--bg-elevated);border-radius:3px;overflow:hidden;">
-    <div style="height:100%;border-radius:3px;width:{{ [pct, 100] | min }}%;
-                background:{% if pct > crit %}var(--red){% elif pct > warn %}var(--orange){% else %}var(--accent){% endif %};
-                transition:width 0.4s;"></div>
-  </div>
-  <span style="font-size:0.7rem;color:var(--text-muted);font-family:ui-monospace,monospace;width:3rem;text-align:right;flex-shrink:0;">{{ "%.1f" | format(pct) }}%</span>
+{% set top = rows[0] %}
+<div class="w-verdict">
+  Busiest: <strong>{{ top.label }}</strong> on {{ top.host_name }},
+  {% if top.usage.cores is not none %}{{ cores(top.usage.cores) }}
+  {%- if top.usage.cpus %} of {{ top.usage.cpus }} cores{% else %} cores{% endif %}{% else %}no CPU reading yet{% endif %}.
 </div>
-{% endmacro %}
-<div style="display:grid;gap:0.6rem;">
+<table class="w-table">
+  <colgroup><col><col class="w-wide-only" style="width:7.5rem"><col class="w-barcol"><col class="w-barcol"></colgroup>
+  <thead><tr><th>Container</th><th class="w-wide-only w-text">Host</th>
+    <th title="Cores, averaged over the last {{ minutes }} minutes; bars: the last {{ hours }}h">CPU<span class="w-wide-only"> · last {{ hours }}h</span></th>
+    <th title="In use now; bars: the last {{ hours }}h">Memory<span class="w-wide-only"> · last {{ hours }}h</span></th></tr></thead>
+  <tbody>
   {% for r in rows %}
-  {% set c = r.c %}
-  <div>
-    <div style="font-size:0.8rem;font-weight:500;white-space:nowrap;overflow:hidden;text-overflow:ellipsis;margin-bottom:0.25rem;">
-      {{ c.name }}{% if multi_host %}<span style="color:var(--text-dim);font-weight:normal;"> · {{ r.host_name }}</span>{% endif %}
-    </div>
-    {% if c.cpu_pct is not none %}{{ bar("CPU", c.cpu_pct, 50, 80) }}{% endif %}
-    {% if c.mem_pct is not none %}{{ bar("MEM", c.mem_pct, 70, 90) }}{% endif %}
-  </div>
+  {% set u = r.usage %}
+  <tr>
+    <td class="w-name" title="{{ r.c.name }}">
+      {%- if session.user_id %}<a href="/plugins/docker/container/{{ r.c.host_id }}/{{ r.c.name }}">{{ r.label }}</a>{% else %}{{ r.label }}{% endif -%}
+    </td>
+    <td class="w-wide-only w-text" title="{{ r.host_name }}">{{ r.host_name }}</td>
+    <td class="tone-{{ (u.cpu_grade or status_class(u.cpu_share, 'cpu')) if u.cpus and u.cores is not none else 'dim' }}"
+        {% if u.cpus and u.cores is not none %}title="{{ '%.1f'|format(u.cpu_share) }}% of {{ r.host_name }}'s {{ u.cpus }} cores{% if u.cpu_usual %}; {{ u.cpu_usual }}{% endif %}"{% endif %}>
+      {{- bars(r.cpu_bars, (cores(r.cpu_bars.peak) ~ " cores") if r.cpu_bars.peak is not none else "", cores(r.cpu_bars.roof) ~ " cores") -}}
+      <span class="w-fig">{% if u.cores is none %}—{% else %}{{ cores(u.cores) }} cores{% endif %}</span>
+    </td>
+    <td class="tone-{{ (u.mem_grade or status_class(u.mem_pct, 'mem')) if u.mem_of and u.mem_bytes is not none else 'dim' }}"
+        {% if u.mem_pct is not none and u.mem_of and u.mem_bytes is not none %}title="{{ '%.1f'|format(u.mem_pct) }}% of {{ u.mem_of }}{% if u.mem_usual %}, {{ u.mem_usual }}{% endif %}"{% endif %}>
+      {{- bars(r.mem_bars, r.mem_bars.peak|bytes if r.mem_bars.peak is not none else "", r.mem_bars.roof|bytes) -}}
+      <span class="w-fig">{% if u.mem_bytes is none %}—{% else %}{{ u.mem_bytes|bytes }}{% endif %}</span>
+    </td>
+  </tr>
   {% endfor %}
-</div>
+  </tbody>
+</table>
+<div class="w-note w-wide-only">Bars: {{ (hours * 60 // 24) }}-minute slices, oldest left; each row up to its own peak, at least {{ cores(cpu_floor) }} core{{ 's' if cpu_floor != 1 }} and {{ mem_floor|bytes }}.</div>
 {% endif %}

plugins/host_agent/__init__.py

@@ -11,13 +11,110 @@ _app: "Quart | None" = None
 def setup(app: "Quart") -> None:
     global _app
     _app = app
+    # A host's CPU, memory and load per core learn their normal (#5742), each
+    # with its floor from Settings → Thresholds. Parts don't: a host's CPU has
+    # a normal, each of its cores doesn't. Disks stay on their fixed cutoffs:
+    # a nearly full disk is a failure whatever its history (System 182).
+    from steward.core.norms import floor_setting, learn_normal
+    learn_normal("host_agent", "cpu_pct", floor_setting(app, "host_cpu_pct"), parts=False)
+    learn_normal("host_agent", "mem_used_pct", floor_setting(app, "host_mem_pct"), parts=False)
+    learn_normal("host_agent", "load_pct", floor_setting(app, "host_load_pct"), parts=False)
+    # "Agent offline" speaks for everything the agent carries: while it is
+    # firing, the host's other rules going silent isn't emailed (#5721).
+    from steward.core.alerts import agent_fed, agent_heartbeat
+    from steward.hosts.service import agent_stale_after_seconds
+    agent_heartbeat("host_agent", "seconds_since_seen", lambda: agent_stale_after_seconds(app.config))
+    agent_fed("host_agent", lambda resource: resource.split(":", 1)[0])
+    # A reboot the agent announced is silence with a reason, for a while (#5756).
+    from datetime import datetime, timezone
+    from steward.core.alerts import agents_away
+    from .downtime import hosts_away, restart_grace_seconds
+    agents_away(lambda session: hosts_away(session, datetime.now(timezone.utc),
+                                           restart_grace_seconds(app.config)))
+    from steward.core.cleanup import prune_by_age
+    from .models import HostDowntime
+    prune_by_age(HostDowntime.started_at)
+
+
+def get_audit_entities() -> dict:
+    """Audit rows about a maintenance plan name it and link to the plans page."""
+    from .models import HostMaintenancePlan
+    return {"maintenance_plan": (HostMaintenancePlan, "/plugins/host_agent/maintenance")}
 
 
 def get_scheduled_tasks() -> list:
-    from .scheduler import make_task
-    return [make_task(_app)]
+    from .scheduler import make_maintenance_task, make_sensor_task, make_task
+    return [make_task(_app), make_maintenance_task(_app), make_sensor_task(_app)]
 
 
 def get_blueprint():
     from .routes import host_agent_bp
     return host_agent_bp
+
+
+def register_mcp_tools(register) -> None:
+    from .mcp_tools import register_tools
+    register_tools(register)
+
+
+async def mcp_overview(db) -> list[dict]:
+    from .mcp_tools import mcp_overview as findings
+    return await findings(db)
+
+
+def get_alert_metrics() -> dict[str, list[str]]:
+    """Host metrics ingest records for alert rules (resource = the host name;
+    disk_used_pct per mount, resource "<host>:<mount>"). seconds_since_seen
+    comes from the scheduler, not the agent: "agent offline" is > 180.
+    clock_off_secs is how far the host's clock is from Steward's, either way
+    (agent ≥ 1.12.0): "clock drifting" is > 30. load_pct is load_1m as a %
+    of the host's cores. inodes_used_pct is per mount (agent ≥ 1.13.0), and
+    disk_used_pct_worst is the fullest mount by space or inodes.
+    mounts_stalled counts the mounts that didn't answer the agent in time
+    (agent ≥ 1.14.0): "an NFS share stopped" is > 0. agent_rss_bytes and
+    agent_cpu_pct are the agent's own memory and CPU (agent ≥ 1.15.0): a
+    healthy agent sits in tens of MB and a few %. updates_pending and
+    updates_security are what the package manager has waiting, and
+    reboot_required is 1 when the host needs a reboot (agent ≥ 1.16.0):
+    "security updates waiting" is updates_security > 0. maintenance_failed
+    comes from the scheduler (#5768): 1 when the host's latest maintenance job
+    failed or timed out, 0 once one succeeds: "maintenance failed" is > 0.
+    temp_c_max is the host's hottest real thermometer, in °C (#5787); sensors
+    with nothing wired to them are left out (#5759, #5783): "running hot" is
+    > 85. temp_c is one sensor, resource "<host>:temp:<label>". net_rx_bps /
+    net_tx_bps and disk_read_bps / disk_write_bps are bytes per second, for
+    the whole host ("<host>") or one interface ("<host>:net:<iface>") or disk
+    ("<host>:diskio:<device>").
+
+    Drive health (agent ≥ 1.20.0, #5789), read hourly with smartctl by the
+    root helper and judged in drives.py, per drive on "<host>:drive:<device>":
+    drive_failing is 1 when the drive fails its own self-assessment, raises
+    an NVMe critical warning or runs out of spare blocks; drive_life_used_pct
+    is how much of its rated life it has used; drive_bad_sectors counts its
+    media errors or reallocated / pending / uncorrectable sectors;
+    drive_temp_c is its own temperature. Per host, for covering rules:
+    drives_failing (how many) and drive_life_used_max_pct (the most worn):
+    "drive failing" is drives_failing > 0, "drive wearing out" is
+    drive_life_used_max_pct > 90.
+
+    Recorded but not offered (NOT_ALERTABLE): disk_used_bytes and
+    disk_total_bytes, where disk_used_pct is the figure to watch (and docker
+    uses disk_total_bytes for something else), and mem_cached_bytes /
+    mem_buffers_bytes, parts of what mem_available_bytes already counts."""
+    from .drives import DRIVE_METRICS
+    return {"host_agent": [
+        "cpu_pct", "mem_used_pct", "mem_available_bytes", "swap_used_bytes",
+        "disk_used_pct_worst", "disk_used_pct", "inodes_used_pct", "load_1m", "load_5m", "load_15m",
+        "load_pct",
+        "uptime_secs", "seconds_since_seen", "clock_off_secs", "mounts_stalled",
+        "agent_rss_bytes", "agent_cpu_pct",
+        "updates_pending", "updates_security", "reboot_required", "maintenance_failed",
+        "temp_c_max", "temp_c", "net_rx_bps", "net_tx_bps", "disk_read_bps", "disk_write_bps",
+        *DRIVE_METRICS,
+    ]}
+
+
+# Metrics host_agent records that no rule is offered, each for the reason
+# get_alert_metrics gives; tests/core/test_metric_catalog.py holds every other
+# recorded name to the catalog.
+NOT_ALERTABLE = frozenset({"disk_used_bytes", "disk_total_bytes", "mem_cached_bytes", "mem_buffers_bytes"})

plugins/host_agent/cadence.py

@@ -0,0 +1,309 @@
+# plugins/host_agent/cadence.py
+"""A group's maintenance cadence (#5768): every week, its hosts updated one at
+a time, each rebooting only when its update needs it (or as the plan says).
+
+A plan names an inventory group, a weekday and a time in its time zone, and a
+window. At that time a run starts with the group's hosts in a fixed order. Each
+host gets a maintenance job (maintenance.py); the next host starts only when
+the last one is back in service: its agent posting, its swarm node ready and
+active, every replicated service at its replicas. A job that fails, or a host
+that doesn't come back into service, stops the run, and the host's
+maintenance_failed figure goes to 1 for an alert rule to catch. No host is
+started after the window; the run then stops and says which hosts it didn't
+reach.
+
+The decisions are model-free here (plans and runs are anything with the
+attributes); the database side is below them, and the plugin's minute tick
+drives it. Everything is on the rows, so a restart of Steward, or the reboot of
+the host it runs on, picks up where it was.
+"""
+from __future__ import annotations
+
+import json
+from datetime import datetime, time, timedelta, timezone
+
+from . import maintenance
+
+WEEKDAYS = ("Monday", "Tuesday", "Wednesday", "Thursday", "Friday", "Saturday", "Sunday")
+# How long a host may take to be back in service after its job finished.
+SETTLE_LIMIT = timedelta(minutes=30)
+# An agent heard from this recently is posting.
+POSTING_WITHIN = timedelta(minutes=3)
+RUN_STATES = ("running", "done", "failed", "stopped")
+
+
+def _zone(name: str):
+    from steward.core.formatting import zone
+    return zone(name)
+
+
+def parse_at(at_time: str) -> time:
+    h, m = (int(x) for x in at_time.split(":", 1))
+    if not (0 <= h < 24 and 0 <= m < 60):
+        raise ValueError(at_time)
+    return time(h, m)
+
+
+def last_slot(plan, now: datetime) -> datetime:
+    """The most recent scheduled time at or before `now`, in UTC."""
+    tz = _zone(plan.time_zone)
+    local = now.astimezone(tz)
+    day = local.date() - timedelta(days=(local.weekday() - plan.weekday) % 7)
+    slot = datetime.combine(day, parse_at(plan.at_time), tzinfo=tz)
+    if slot > local:
+        slot -= timedelta(days=7)
+    return slot.astimezone(timezone.utc)
+
+
+def next_slot(plan, now: datetime) -> datetime:
+    return last_slot(plan, now) + timedelta(days=7)
+
+
+def is_due(plan, now: datetime) -> datetime | None:
+    """The slot to fire now, or None. A slot fires once (last_slot_at), and
+    only while its window is still open: a Steward that was down over the slot
+    starts late, not a day late."""
+    if not plan.enabled:
+        return None
+    slot = last_slot(plan, now)
+    since = plan.last_slot_at or plan.created_at
+    if since is not None and slot <= since:
+        return None
+    if now - slot >= timedelta(minutes=plan.window_minutes):
+        return None
+    return slot
+
+
+def back_in_service(*, last_seen_at, node, services, job_finished_at, now: datetime,
+                    expect_active: bool = True) -> str | None:
+    """None when the host is back in service, else why not yet. `node` is the
+    freshest swarm-node row for the job's node (None outside a swarm, or when
+    the job has no node); `services` the freshest row per swarm service. Only
+    rows reported after the job finished count, so a stale "all fine" from
+    before the reboot never passes. `expect_active` is False for a node the
+    job didn't drain (the operator keeps it drained): it need only be ready."""
+    if last_seen_at is None or now - last_seen_at > POSTING_WITHIN:
+        return "its agent isn't posting yet"
+    if node is None:
+        return None
+    if node.updated_at < job_finished_at:
+        return "no swarm report since it came back"
+    if node.status != "ready" or (expect_active and node.availability != "active"):
+        return f"its swarm node is {node.status}/{node.availability}"
+    fresh = [s for s in services if s.updated_at >= job_finished_at]
+    if not fresh:
+        return "no swarm report since it came back"
+    short = sorted(s.service_name for s in fresh
+                   if s.mode == "replicated" and s.running < s.desired)
+    if short:
+        return "services below their replicas: " + ", ".join(short[:5])
+    return None
+
+
+def next_move(run, plan, job, back: str | None, now: datetime) -> tuple[str, str | None]:
+    """What a running run does now: ("start", None) the current host's job;
+    ("wait", None); ("next", None) on to the next host; ("settling", why);
+    ("fail", why); ("stop", why) the window ended; ("done", None)."""
+    hosts = json.loads(run.host_ids)
+    if run.position >= len(hosts):
+        return "done", None
+    if job is None:
+        if now >= run.started_at + timedelta(minutes=plan.window_minutes):
+            left = len(hosts) - run.position
+            return "stop", f"the window ended with {left} host{'s' if left != 1 else ''} not started"
+        return "start", None
+    if job.state in maintenance.ACTIVE:
+        return "wait", None
+    if job.state != "done":
+        return "fail", job.detail or maintenance.LABELS[job.state]
+    if back is None:
+        return "next", None
+    if run.settle_since is not None and now - run.settle_since > SETTLE_LIMIT:
+        return "fail", f"not back in service within {int(SETTLE_LIMIT.total_seconds() // 60)} min: {back}"
+    return "settling", back
+
+
+def describe(plan) -> str:
+    """"Sundays at 03:00 (Europe/London), up to 4 h" for the page."""
+    hours = plan.window_minutes / 60
+    span = f"{hours:g} h" if plan.window_minutes % 60 == 0 else f"{plan.window_minutes} min"
+    return f"{WEEKDAYS[plan.weekday]}s at {plan.at_time} ({plan.time_zone}), up to {span}"
+
+
+# ── applied to the database: the minute tick, Run now, the page ──────────────
+
+async def group_hosts(session, group_id: str) -> list:
+    """The group's hosts (through their linked Ansible targets), by name."""
+    from sqlalchemy import select
+    from steward.models.ansible_inventory import AnsibleTarget, ansible_target_groups
+    from steward.models.hosts import Host
+    rows = (await session.execute(
+        select(Host).join(AnsibleTarget, AnsibleTarget.host_id == Host.id)
+        .join(ansible_target_groups, ansible_target_groups.c.target_id == AnsibleTarget.id)
+        .where(ansible_target_groups.c.group_id == group_id))).scalars().all()
+    return sorted({h.id: h for h in rows}.values(), key=lambda h: h.name.lower())
+
+
+async def start_run(session, plan, now: datetime, manual: bool = False):
+    """A run of the plan with the group's hosts in order. A host whose agent
+    has no maintenance helper is passed over, and the run says so."""
+    from sqlalchemy import select
+    from .models import HostAgentRegistration, HostMaintenanceRun
+    hosts = await group_hosts(session, plan.group_id)
+    regs = {r.host_id: r for r in (await session.execute(select(HostAgentRegistration).where(
+        HostAgentRegistration.host_id.in_([h.id for h in hosts])))).scalars()} if hosts else {}
+    order, skipped = [], []
+    for h in hosts:
+        reg = regs.get(h.id)
+        if reg is None:
+            skipped.append({"host_id": h.id, "reason": "no agent"})
+        elif not reg.maint_helper:
+            skipped.append({"host_id": h.id, "reason": "no maintenance helper: run Update agent"})
+        else:
+            order.append(h.id)
+    run = HostMaintenanceRun(plan_id=plan.id, state="running", host_ids=json.dumps(order),
+                             position=0, skipped=json.dumps(skipped), manual=manual,
+                             started_at=now)
+    session.add(run)
+    await session.flush()
+    return run
+
+
+async def _back(session, host, job, reg, now: datetime) -> str | None:
+    from sqlalchemy import select
+    node = None
+    services: list = []
+    if job.node_id:
+        models = maintenance.docker_models()
+        if models is not None:
+            DockerSwarmNode, DockerSwarmService = models.DockerSwarmNode, models.DockerSwarmService
+            node = (await session.execute(select(DockerSwarmNode).where(
+                DockerSwarmNode.node_id == job.node_id)
+                .order_by(DockerSwarmNode.updated_at.desc()).limit(1))).scalar_one_or_none()
+            freshest: dict = {}
+            for s in (await session.execute(select(DockerSwarmService))).scalars():
+                if s.service_name not in freshest or s.updated_at > freshest[s.service_name].updated_at:
+                    freshest[s.service_name] = s
+            services = list(freshest.values())
+    return back_in_service(last_seen_at=reg.last_seen_at if reg else None, node=node,
+                           services=services, job_finished_at=job.finished_at or now, now=now,
+                           expect_active=bool(job.drain_host_id))
+
+
+async def advance(session, run, plan, now: datetime) -> str:
+    """Move one running run on as far as it can go now; returns its state."""
+    from sqlalchemy import select
+    from steward.models.hosts import Host
+    from .models import HostAgentRegistration, HostMaintenanceJob
+    for _ in range(64):                             # a host per pass; bounded
+        hosts = json.loads(run.host_ids)
+        host_id = hosts[run.position] if run.position < len(hosts) else None
+        job = host = reg = None
+        if host_id is not None:
+            host = await session.get(Host, host_id)
+            reg = (await session.execute(select(HostAgentRegistration).where(
+                HostAgentRegistration.host_id == host_id))).scalar_one_or_none()
+            job = (await session.execute(select(HostMaintenanceJob).where(
+                HostMaintenanceJob.run_id == run.id, HostMaintenanceJob.host_id == host_id)
+                .order_by(HostMaintenanceJob.created_at.desc()).limit(1))).scalar_one_or_none()
+        back = None
+        if job is not None and job.state == "done":
+            back = await _back(session, host, job, reg, now)
+        move, why = next_move(run, plan, job, back, now)
+        if move == "start":
+            if host is None:
+                _skip(run, host_id, "the host is gone")
+                continue
+            try:
+                await maintenance.create_job(session, host, reg, plan.created_by,
+                                             plan.reboot_policy, now, run_id=run.id)
+            except maintenance.CannotStart as exc:
+                _skip(run, host_id, str(exc))
+                continue
+            return run.state
+        if move == "next":
+            run.position += 1
+            run.settle_since = None
+            continue
+        if move == "settling":
+            run.settle_since = run.settle_since or now
+            return run.state
+        if move == "fail":
+            if job is not None and job.state == "done":
+                maintenance._finish(job, "failed", now, why)     # the host never came back into service
+            name = host.name if host else host_id
+            _end(run, "failed", now, f"{name}: {why}")
+        elif move == "stop":
+            _end(run, "stopped", now, why)
+        elif move == "done":
+            _end(run, "done", now, None)
+        return run.state
+    return run.state
+
+
+def _skip(run, host_id: str, reason: str) -> None:
+    skipped = json.loads(run.skipped)
+    skipped.append({"host_id": host_id, "reason": reason})
+    run.skipped = json.dumps(skipped)
+    hosts = json.loads(run.host_ids)
+    hosts.pop(run.position)
+    run.host_ids = json.dumps(hosts)
+
+
+def _end(run, state: str, now: datetime, detail: str | None) -> None:
+    run.state, run.finished_at, run.settle_since = state, now, None
+    if detail:
+        run.detail = detail[:maintenance.DETAIL_MAX]
+
+
+async def tick(session, now: datetime) -> list[str]:
+    """Start due plans and move running runs on; returns lines worth logging."""
+    from sqlalchemy import select
+    from .models import HostMaintenancePlan, HostMaintenanceRun
+    said = []
+    running = {r.plan_id: r for r in (await session.execute(select(HostMaintenanceRun).where(
+        HostMaintenanceRun.state == "running"))).scalars()}
+    for plan in (await session.execute(select(HostMaintenancePlan))).scalars():
+        slot = is_due(plan, now)
+        if slot is None:
+            continue
+        plan.last_slot_at = slot
+        if plan.id in running:
+            said.append(f"plan {plan.name!r}: the {slot:%Y-%m-%d %H:%M}Z slot came with a run still going")
+            continue
+        running[plan.id] = await start_run(session, plan, now)
+        said.append(f"plan {plan.name!r}: run started")
+    for run in running.values():
+        plan = await session.get(HostMaintenancePlan, run.plan_id)
+        if plan is None:
+            continue
+        before = run.state
+        after = await advance(session, run, plan, now)
+        if after != before:
+            said.append(f"plan {plan.name!r}: run {after}" + (f" ({run.detail})" if run.detail else ""))
+    return said
+
+
+async def failed_hosts(session, now: datetime) -> dict[str, float]:
+    """{host name: 1 or 0} for every host with an agent: whether its latest
+    finished maintenance job failed or timed out, for the maintenance_failed
+    figure an alert rule watches. A host with no job yet reads 0, so a rule
+    on it reports rather than sitting at "no data". A later job that
+    succeeds puts it back to 0."""
+    from sqlalchemy import select
+    from steward.models.hosts import Host
+    from .models import HostAgentRegistration, HostMaintenanceJob
+    agents = (await session.execute(
+        select(Host.name).join(HostAgentRegistration, HostAgentRegistration.host_id == Host.id))).scalars()
+    out = {name: 0.0 for name in agents}
+    rows = (await session.execute(
+        select(Host.name, HostMaintenanceJob.state, HostMaintenanceJob.finished_at)
+        .join(Host, Host.id == HostMaintenanceJob.host_id)
+        .where(HostMaintenanceJob.state.in_(maintenance.FINISHED),
+               HostMaintenanceJob.finished_at >= now - timedelta(days=30)))).all()
+    latest: dict = {}
+    for name, state, finished in rows:
+        if name not in latest or finished > latest[name][1]:
+            latest[name] = (state, finished)
+    out.update({name: 0.0 if state == "done" else 1.0 for name, (state, _) in latest.items()})
+    return out

plugins/host_agent/downtime.py

@@ -0,0 +1,160 @@
+# plugins/host_agent/downtime.py
+"""What a host's agent says happened to it while it was away (#5756).
+
+An agent ≥ 1.14.0 tells Steward when it stops (and whether its host is going
+down with it), when it starts (and when its host booted), and how long it
+couldn't reach Steward. Each becomes a HostDowntime row, so a reboot reads as
+a reboot with its length, and "agent offline" is left for silence nobody
+explained. A post can be retried, so every event is applied idempotently.
+"""
+from __future__ import annotations
+
+from datetime import datetime, timedelta, timezone
+
+from sqlalchemy import select
+
+from .models import HostDowntime
+
+# A stop the agent announced: the silence after it is explained, for a while.
+ANNOUNCED = ("reboot", "agent_stop")
+
+LABELS = {
+    "reboot": "Rebooted",
+    "agent_stop": "Agent stopped",
+    "unclean": "Restarted without shutting down",
+    "lost_contact": "Lost contact",
+}
+
+
+def restart_grace_seconds(config) -> int:
+    """How long an announced stop explains the agent's silence (Settings →
+    Plugins → host_agent, default 10 minutes): past it, agent offline fires."""
+    return int(((config.get("PLUGINS") or {}).get("host_agent") or {})
+               .get("reboot_grace_seconds", 600))
+
+
+def _at(value) -> datetime | None:
+    if not isinstance(value, str):
+        return None
+    try:
+        got = datetime.fromisoformat(value.replace("Z", "+00:00"))
+    except ValueError:
+        return None
+    return got if got.tzinfo else got.replace(tzinfo=timezone.utc)
+
+
+def _when(event: dict) -> datetime | None:
+    return _at(event.get("at") or event.get("from"))
+
+
+async def record_events(session, host_id: str, events, last_seen_at: datetime | None) -> int:
+    """Apply one post's events to the host's downtime, oldest first; returns
+    how many rows were added or changed. `last_seen_at` is when the agent was
+    last heard before this post: a boot after it that nobody announced is a
+    restart without a shutdown."""
+    if not isinstance(events, list):
+        return 0
+    timed = sorted(((t, e) for e in events if isinstance(e, dict)
+                    for t in [_when(e)] if t is not None), key=lambda te: te[0])
+    changed = 0
+    for at, event in timed:
+        kind = event.get("kind")
+        if kind == "stopping":
+            changed += await _stopping(session, host_id, at, event.get("host_shutdown"))
+        elif kind == "started":
+            changed += await _started(session, host_id, at, _at(event.get("boot_at")), last_seen_at)
+        elif kind == "lost_contact":
+            changed += await _lost_contact(session, host_id, at, _at(event.get("to")))
+    return changed
+
+
+async def _stopping(session, host_id: str, at: datetime, host_shutdown) -> int:
+    seen = (await session.execute(select(HostDowntime.id).where(
+        HostDowntime.host_id == host_id, HostDowntime.kind.in_(ANNOUNCED),
+        HostDowntime.started_at == at))).first()
+    if seen:
+        return 0
+    session.add(HostDowntime(host_id=host_id, kind="reboot" if host_shutdown else "agent_stop",
+                             started_at=at))
+    return 1
+
+
+async def _started(session, host_id: str, at: datetime, boot_at: datetime | None,
+                   last_seen_at: datetime | None) -> int:
+    stop = (await session.execute(select(HostDowntime).where(
+        HostDowntime.host_id == host_id, HostDowntime.kind.in_(ANNOUNCED),
+        HostDowntime.ended_at.is_(None), HostDowntime.started_at <= at)
+        .order_by(HostDowntime.started_at.desc()).limit(1))).scalar_one_or_none()
+    if stop is not None:
+        stop.ended_at = at
+        # The boot time settles what the stop was: the host came back from a
+        # boot after it, or it stayed up while only the agent restarted.
+        if boot_at is not None:
+            stop.kind = "reboot" if boot_at > stop.started_at else "agent_stop"
+        return 1
+    if boot_at is None or last_seen_at is None or boot_at <= last_seen_at:
+        return 0
+    seen = (await session.execute(select(HostDowntime.id).where(
+        HostDowntime.host_id == host_id, HostDowntime.kind == "unclean",
+        HostDowntime.ended_at == at))).first()
+    if seen:
+        return 0
+    session.add(HostDowntime(host_id=host_id, kind="unclean", started_at=last_seen_at, ended_at=at))
+    return 1
+
+
+async def _lost_contact(session, host_id: str, since: datetime, until: datetime | None) -> int:
+    if until is None or until <= since:
+        return 0
+    row = (await session.execute(select(HostDowntime).where(
+        HostDowntime.host_id == host_id, HostDowntime.kind == "lost_contact",
+        HostDowntime.started_at == since))).scalar_one_or_none()
+    if row is None:
+        session.add(HostDowntime(host_id=host_id, kind="lost_contact", started_at=since, ended_at=until))
+        return 1
+    if row.ended_at is None or until > row.ended_at:
+        row.ended_at = until
+        return 1
+    return 0
+
+
+async def announced_stops(session, now: datetime, grace_seconds: int) -> dict[str, datetime]:
+    """{host id: when the silence its agent announced stops being explained}
+    for every announced stop still going and inside its grace."""
+    rows = (await session.execute(select(HostDowntime.host_id, HostDowntime.started_at).where(
+        HostDowntime.kind.in_(ANNOUNCED), HostDowntime.ended_at.is_(None),
+        HostDowntime.started_at > now - timedelta(seconds=grace_seconds)))).all()
+    out: dict[str, datetime] = {}
+    for host_id, started in rows:
+        until = started + timedelta(seconds=grace_seconds)
+        out[host_id] = max(out.get(host_id, until), until)
+    return out
+
+
+async def hosts_away(session, now: datetime, grace_seconds: int) -> set[str]:
+    """Names of the hosts in announced_stops: their silence has a reason."""
+    from steward.models.hosts import Host
+    ids = list(await announced_stops(session, now, grace_seconds))
+    if not ids:
+        return set()
+    return set((await session.execute(select(Host.name).where(Host.id.in_(ids)))).scalars())
+
+
+def describe(row: HostDowntime, now: datetime) -> dict:
+    """One downtime row as the host page shows it."""
+    label = LABELS.get(row.kind, row.kind)
+    if row.kind == "agent_stop" and row.ended_at is not None:
+        label = "Agent restarted"
+    end = row.ended_at or now
+    return {"kind": row.kind, "label": label, "started_at": row.started_at,
+            "ended_at": row.ended_at, "ongoing": row.ended_at is None,
+            "seconds": max(0.0, (end - row.started_at).total_seconds())}
+
+
+async def recent(session, host_id: str, now: datetime, days: int = 7, limit: int = 6) -> list[dict]:
+    """The host's downtime in the last `days`, newest first."""
+    rows = (await session.execute(select(HostDowntime).where(
+        HostDowntime.host_id == host_id,
+        HostDowntime.started_at > now - timedelta(days=days))
+        .order_by(HostDowntime.started_at.desc()).limit(limit))).scalars().all()
+    return [describe(r, now) for r in rows]

plugins/host_agent/drives.py

@@ -0,0 +1,116 @@
+"""Each drive's health, judged from what the root helper read with smartctl
+(#5789). The helper reports figures, not verdicts, so how a drive is judged
+can change here without an agent update.
+
+A drive is:
+- "failing" when its own self-assessment fails (SMART overall-health), an NVMe
+  drive raises a critical warning, or its spare blocks fall under threshold;
+- "worn" when it has used WORN_PCT of its rated life;
+- "watch" when it has bad sectors or media errors but otherwise passes: old
+  disks carry a few for years, so a count alone isn't a failure;
+- "ok", or "unavailable" when it has no SMART to read (a virtual disk, a
+  bridge that hides it), or "unreadable" when smartctl couldn't read it.
+
+Metrics, per drive on "<host>:drive:<device>": drive_failing (1/0),
+drive_life_used_pct, drive_bad_sectors, drive_temp_c. Per host, for covering
+rules: drives_failing (how many) and drive_life_used_max_pct.
+"""
+from __future__ import annotations
+
+WORN_PCT = 90
+DRIVE_METRICS = ("drive_failing", "drive_life_used_pct", "drive_bad_sectors", "drive_temp_c",
+                 "drives_failing", "drive_life_used_max_pct")
+# What a drive's own figures count as bad sectors or media errors, whichever
+# its kind reports: NVMe media errors, ATA reallocated / pending / offline
+# uncorrectable sectors, the SCSI grown defect list.
+_BAD = ("media_errors", "reallocated", "pending", "offline_uncorrectable", "grown_defects")
+_ORDER = {"failing": 0, "unreadable": 1, "worn": 2, "watch": 3, "ok": 4, "unavailable": 5}
+
+
+def _num(v):
+    return v if isinstance(v, (int, float)) and not isinstance(v, bool) else None
+
+
+def bad_sectors(d: dict) -> int | None:
+    counts = [_num(d.get(k)) for k in _BAD]
+    counts = [c for c in counts if c is not None]
+    return int(sum(counts)) if counts else None
+
+
+def failing_reasons(d: dict) -> list[str]:
+    why = []
+    if d.get("passed") is False:
+        why.append("its SMART self-assessment failed")
+    if _num(d.get("critical_warning")):
+        why.append("critical warning %#x" % int(d["critical_warning"]))
+    spare, floor = _num(d.get("spare_pct")), _num(d.get("spare_threshold_pct"))
+    if spare is not None and floor is not None and spare < floor:
+        why.append(f"spare blocks {spare:.0f}% under the {floor:.0f}% threshold")
+    return why
+
+
+def judge(d: dict) -> dict:
+    """One drive as the host page shows it: its figures plus `state` and `why`."""
+    out = dict(d)
+    if d.get("error"):
+        out.update(state="unreadable", why=str(d["error"]))
+    elif d.get("available") is False:
+        out.update(state="unavailable", why="no SMART data (a virtual disk, or SMART is off)")
+    elif failing_reasons(d):
+        out.update(state="failing", why="; ".join(failing_reasons(d)))
+    elif (_num(d.get("life_used_pct")) or 0) >= WORN_PCT:
+        out.update(state="worn", why=f"{d['life_used_pct']:.0f}% of its rated life used")
+    elif bad_sectors(d):
+        out.update(state="watch", why=f"{bad_sectors(d)} bad sectors or media errors")
+    else:
+        out.update(state="ok", why="")
+    out["bad_sectors"] = bad_sectors(d)
+    return out
+
+
+def drives_of(report) -> list[dict]:
+    """The report's drives, judged, what needs attention first. A name seen
+    twice gets an ordinal, so two drives never share one series: agent
+    1.20.0 named every disk behind a RAID controller "0" (#5806)."""
+    drives = report.get("drives") if isinstance(report, dict) else None
+    judged, seen = [], {}
+    for d in drives or []:
+        if not isinstance(d, dict) or not d.get("device"):
+            continue
+        d = judge(d)
+        name = str(d["device"])
+        seen[name] = seen.get(name, 0) + 1
+        if seen[name] > 1:
+            d["device"] = f"{name}#{seen[name]}"
+        judged.append(d)
+    return sorted(judged, key=lambda d: (_ORDER[d["state"]], str(d["device"])))
+
+
+def metric_rows(report, host_name: str) -> list[tuple[str, str, float]]:
+    """(metric, resource, value) for every figure a drive has, and the host's
+    two roll-ups once any drive was read."""
+    rows: list[tuple[str, str, float]] = []
+    failing = 0
+    life_max: float | None = None
+    read = 0
+    for d in drives_of(report):
+        if d["state"] in ("unreadable", "unavailable"):
+            continue
+        read += 1
+        res = f"{host_name}:drive:{d['device']}"
+        is_failing = d["state"] == "failing"
+        failing += is_failing
+        rows.append(("drive_failing", res, 1.0 if is_failing else 0.0))
+        life = _num(d.get("life_used_pct"))
+        if life is not None:
+            rows.append(("drive_life_used_pct", res, float(life)))
+            life_max = life if life_max is None else max(life_max, life)
+        if d["bad_sectors"] is not None:
+            rows.append(("drive_bad_sectors", res, float(d["bad_sectors"])))
+        if _num(d.get("temp_c")) is not None:
+            rows.append(("drive_temp_c", res, float(d["temp_c"])))
+    if read:
+        rows.append(("drives_failing", host_name, float(failing)))
+        if life_max is not None:
+            rows.append(("drive_life_used_max_pct", host_name, float(life_max)))
+    return rows

plugins/host_agent/files/install-deps.sh

@@ -0,0 +1,50 @@
+#!/bin/sh
+# Steward host agent: the packages the agent's helper uses, installed with
+# whichever package manager the host has (#5789). Served by Steward at
+# /plugins/host_agent/files/install-deps.sh and run, as root, by install.sh
+# and the host_agent playbooks: one copy of the list.
+#
+# smartmontools gives the helper smartctl, which reads each drive's health.
+# A package that can't be installed is reported, never fatal: the agent runs
+# without it and the host page says what is missing.
+#
+# Recommended / weak dependencies are skipped: on Debian, smartmontools
+# recommends a mail server.
+
+PACKAGES="smartmontools"
+
+have() { command -v "$1" >/dev/null 2>&1; }
+
+install_with() {
+    if have apt-get; then
+        export DEBIAN_FRONTEND=noninteractive
+        # A stale package list can't find the package: refresh once and retry.
+        apt-get install -y --no-install-recommends "$@" \
+            || { apt-get update && apt-get install -y --no-install-recommends "$@"; }
+    elif have dnf; then
+        dnf -y install --setopt=install_weak_deps=False "$@"
+    elif have yum; then
+        yum -y install "$@"
+    elif have zypper; then
+        zypper --non-interactive install --no-recommends "$@"
+    elif have pacman; then
+        pacman -S --needed --noconfirm "$@"
+    elif have apk; then
+        apk add --no-cache "$@"
+    else
+        echo "no package manager this script knows (apt, dnf, yum, zypper, pacman, apk)"
+        return 1
+    fi
+}
+
+missing=""
+have smartctl || missing="$missing smartmontools"
+[ -n "$missing" ] || { echo "dependencies present: $PACKAGES"; exit 0; }
+
+# shellcheck disable=SC2086  # one word per package
+if install_with $missing; then
+    echo "installed:$missing"
+else
+    echo "WARNING: could not install$missing; drive health will read \"smartctl not installed\" until it is"
+fi
+exit 0

plugins/host_agent/files/maint.py

@@ -0,0 +1,387 @@
+#!/usr/bin/env python3
+"""Steward maintenance helper (#5766): runs as root, started by steward-maint.path
+when the host agent leaves a request, and does only what ACTIONS lists.
+
+The agent (an unprivileged user under NoNewPrivileges) can't update or reboot
+its host itself. It writes `<id>.json` = {"id", "action", ["node"]} into
+REQUESTS, the one directory it may write. This helper reads each request
+without following links, refuses anything outside the list, deletes it, does
+the action, and writes the outcome to RESULTS (root's own directory, which the
+agent reads back and reports to Steward). A stolen agent token can therefore
+make its host update, reboot or drain, and nothing else.
+
+One action only reads: `drives` (#5789) asks smartctl for each disk's health,
+which needs root, and writes a summary to DRIVES, never a result: it is not a
+maintenance job.
+
+Stdlib only; served by Steward at /plugins/host_agent/files/maint.py.
+"""
+from __future__ import annotations
+
+import json
+import os
+import re
+import shutil
+import subprocess
+import sys
+import time
+from datetime import datetime, timezone
+
+REQUESTS = "/var/lib/steward-agent/requests"
+RESULTS = "/var/lib/steward-maint/results"
+DRIVES = "/var/lib/steward-maint/drives.json"
+ACTIONS = ("update", "reboot", "drain", "undrain", "drives")
+REQUEST_MAX_BYTES = 4096
+RESULTS_KEPT_SECONDS = 7 * 86400
+UPDATE_DEADLINE_SECONDS = 3000
+COMMAND_DEADLINE_SECONDS = 600
+OUTPUT_KEPT_CHARS = 2000
+_ID = re.compile(r"^[A-Za-z0-9-]{1,64}$")
+_NODE = re.compile(r"^[a-z0-9]{25}$")
+
+
+def now_iso() -> str:
+    return datetime.now(timezone.utc).isoformat()
+
+
+def read_request(path: str):
+    """The request at `path` as a dict, or None when it is not a small regular
+    file holding a valid request. Never follows a link the agent planted."""
+    try:
+        fd = os.open(path, os.O_RDONLY | os.O_NOFOLLOW | os.O_NONBLOCK)
+    except OSError:
+        return None
+    try:
+        st = os.fstat(fd)
+        if not (st.st_mode & 0o170000 == 0o100000) or st.st_size > REQUEST_MAX_BYTES:
+            return None
+        raw = os.read(fd, REQUEST_MAX_BYTES + 1)
+    finally:
+        os.close(fd)
+    try:
+        req = json.loads(raw.decode("utf-8"))
+    except (UnicodeDecodeError, ValueError):
+        return None
+    return req if valid(req) else None
+
+
+def valid(req) -> bool:
+    """A request names a known action, has a plain id, and a node id only in
+    the swarm's own form."""
+    if not isinstance(req, dict) or req.get("action") not in ACTIONS:
+        return False
+    if not isinstance(req.get("id"), str) or not _ID.match(req["id"]):
+        return False
+    node = req.get("node")
+    return node is None or (isinstance(node, str) and bool(_NODE.match(node)))
+
+
+def run(cmd: list, deadline: int = COMMAND_DEADLINE_SECONDS, env: dict | None = None):
+    """(ok, output tail) of a command, bounded by `deadline`."""
+    try:
+        done = subprocess.run(cmd, capture_output=True, text=True, timeout=deadline,
+                              env=dict(os.environ, LANG="C", LC_ALL="C", **(env or {})))
+    except subprocess.TimeoutExpired:
+        return False, f"{cmd[0]} didn't finish in {deadline} s"
+    except OSError as exc:
+        return False, str(exc)
+    out = (done.stdout or "") + (done.stderr or "")
+    return done.returncode == 0, out[-OUTPUT_KEPT_CHARS:]
+
+
+def reboot_required() -> bool | None:
+    if os.path.exists("/var/run/reboot-required"):
+        return True
+    if shutil.which("apt-get"):
+        return False
+    if shutil.which("needs-restarting"):
+        try:
+            return subprocess.run(["needs-restarting", "-r"], capture_output=True,
+                                  timeout=60).returncode == 1
+        except (OSError, subprocess.TimeoutExpired):
+            return None
+    return None
+
+
+def upgraded_count(output: str) -> int | None:
+    """Packages an apt or dnf run changed, from its summary line."""
+    m = re.search(r"(\d+) upgraded, (\d+) newly installed", output)       # apt
+    if m:
+        return int(m.group(1)) + int(m.group(2))
+    counts = re.findall(r"^\s*(?:Upgrade|Install)\s+(\d+) Packages?", output, re.MULTILINE)  # dnf
+    return sum(int(c) for c in counts) if counts else None
+
+
+def do_update() -> dict:
+    """Upgrade every package, keeping existing config files and leaving
+    service restarts to a reboot: needrestart only lists, so the Docker daemon
+    isn't restarted under its containers mid-upgrade."""
+    if shutil.which("apt-get"):
+        env = {"DEBIAN_FRONTEND": "noninteractive", "NEEDRESTART_MODE": "l"}
+        ok, out = run(["apt-get", "update"], env=env)
+        if not ok:
+            return {"ok": False, "detail": "apt-get update failed: " + out}
+        ok, out = run(["apt-get", "-y", "-o", "Dpkg::Options::=--force-confdef",
+                       "-o", "Dpkg::Options::=--force-confold", "dist-upgrade"],
+                      deadline=UPDATE_DEADLINE_SECONDS, env=env)
+        if not ok:
+            return {"ok": False, "detail": "apt-get dist-upgrade failed: " + out}
+        changed = upgraded_count(out)
+        run(["apt-get", "-y", "autoremove"], env=env)
+    elif shutil.which("dnf"):
+        ok, out = run(["dnf", "-y", "upgrade", "--refresh"], deadline=UPDATE_DEADLINE_SECONDS)
+        if not ok:
+            return {"ok": False, "detail": "dnf upgrade failed: " + out}
+        changed = upgraded_count(out)
+    else:
+        return {"ok": False, "detail": "no apt-get or dnf on this host"}
+    return {"ok": True, "packages": changed, "reboot_required": reboot_required(),
+            "detail": out[-500:]}
+
+
+def own_node() -> str | None:
+    ok, out = run(["docker", "info", "--format", "{{.Swarm.NodeID}}"], deadline=30)
+    node = out.strip() if ok else ""
+    return node if _NODE.match(node) else None
+
+
+def do_node(availability: str, node: str | None) -> dict:
+    """Set a swarm node's availability (drain before a reboot, active after).
+    Only a manager can; a worker's helper says so and Steward asks a manager."""
+    if not shutil.which("docker"):
+        return {"ok": False, "detail": "no docker on this host"}
+    node = node or own_node()
+    if node is None:
+        return {"ok": False, "detail": "this host is not in a swarm"}
+    ok, out = run(["docker", "node", "update", "--availability", availability, node], deadline=60)
+    return {"ok": ok, "node": node, "detail": out.strip()[-500:]}
+
+
+# ── drive health (#5789) ─────────────────────────────────────────────────────
+# smartctl ≥ 7.0 answers in JSON, the same for SATA, SAS and NVMe drives on any
+# distribution; an older one is reported, never parsed as text. The helper
+# finds the drives itself: the request names none.
+
+SMARTCTL_MIN = (7, 0)
+SMARTCTL_DEADLINE_SECONDS = 60
+DRIVES_MAX = 32
+# smartctl's exit status is a bit mask: 1 the command line, 2 the device
+# couldn't be opened. Higher bits describe the disk, and its JSON still holds.
+_SMARTCTL_UNREAD = 0b11
+# ATA attributes counted as bad sectors.
+_ATA_BAD = {5: "reallocated", 197: "pending", 198: "offline_uncorrectable"}
+
+
+def smartctl_version():
+    """(major, minor) of the installed smartctl, or None when there is none."""
+    if not shutil.which("smartctl"):
+        return None
+    ok, out = run(["smartctl", "--version"], deadline=30)
+    m = re.search(r"smartctl (\d+)\.(\d+)", out)
+    return (int(m.group(1)), int(m.group(2))) if m else (0, 0)
+
+
+def smartctl_json(args: list):
+    """(exit status, parsed JSON or None) of one smartctl call."""
+    try:
+        done = subprocess.run(["smartctl", "--json", *args], capture_output=True, text=True,
+                              timeout=SMARTCTL_DEADLINE_SECONDS,
+                              env=dict(os.environ, LANG="C", LC_ALL="C"))
+    except (subprocess.TimeoutExpired, OSError):
+        return None, None
+    try:
+        return done.returncode, json.loads(done.stdout)
+    except ValueError:
+        return done.returncode, None
+
+
+def _int(v):
+    return v if isinstance(v, int) and not isinstance(v, bool) else None
+
+
+def drive_name(name: str, kind) -> str:
+    """A drive's name, unique on its host. Disks behind a RAID controller
+    share one path (/dev/bus/0) and differ only by the slot in their type
+    (megaraid,3; cciss,0), so the slot is part of the name (#5806)."""
+    base = os.path.basename(name)
+    if name.startswith("/dev/bus/"):
+        base = "bus" + base
+    if isinstance(kind, str) and "," in kind:
+        base += "-" + re.sub(r"[^A-Za-z0-9]", "", kind.split("+")[-1])
+    return base
+
+
+def drive_summary(name: str, status, data, kind=None) -> dict:
+    """The few figures Steward reads from one drive's `smartctl -a` JSON."""
+    out = {"device": drive_name(name, kind)}
+    if not isinstance(data, dict) or status is None or status & _SMARTCTL_UNREAD:
+        messages = ((data.get("smartctl") or {}).get("messages") or []) if isinstance(data, dict) else []
+        text = "; ".join(str(m.get("string", "")) for m in messages if isinstance(m, dict))
+        out["error"] = (text or "smartctl couldn't read it")[:200]
+        return out
+    dev = data.get("device") or {}
+    out["protocol"] = str(dev.get("protocol") or "")[:16]
+    out["model"] = str(data.get("model_name") or data.get("scsi_model_name") or "")[:80]
+    cap = (data.get("user_capacity") or {}).get("bytes")
+    if _int(cap):
+        out["capacity_bytes"] = cap
+    support = data.get("smart_support") or {}
+    if support.get("available") is False or support.get("enabled") is False:
+        out["available"] = False             # a virtual disk, or SMART switched off
+        return out
+    passed = (data.get("smart_status") or {}).get("passed")
+    if isinstance(passed, bool):
+        out["passed"] = passed
+    temp = _int((data.get("temperature") or {}).get("current"))
+    if temp is not None:
+        out["temp_c"] = temp
+    hours = _int((data.get("power_on_time") or {}).get("hours"))
+    if hours is not None:
+        out["power_on_hours"] = hours
+    nvme = data.get("nvme_smart_health_information_log") or {}
+    for src, dst in (("critical_warning", "critical_warning"), ("percentage_used", "life_used_pct"),
+                     ("media_errors", "media_errors"), ("available_spare", "spare_pct"),
+                     ("available_spare_threshold", "spare_threshold_pct")):
+        if _int(nvme.get(src)) is not None:
+            out[dst] = nvme[src]
+    for attr in (data.get("ata_smart_attributes") or {}).get("table") or []:
+        key = _ATA_BAD.get(attr.get("id")) if isinstance(attr, dict) else None
+        raw = _int((attr.get("raw") or {}).get("value")) if key else None
+        if raw is not None:
+            out[key] = raw
+    used = _int((data.get("endurance_used") or {}).get("current_percent"))          # ATA SSDs
+    used = used if used is not None else _int(data.get("scsi_percentage_used_endurance_indicator"))
+    if used is not None and "life_used_pct" not in out:
+        out["life_used_pct"] = used
+    if _int(data.get("scsi_grown_defect_list")) is not None:
+        out["grown_defects"] = data["scsi_grown_defect_list"]
+    return out
+
+
+def do_drives() -> dict:
+    """Each drive's health, as {"checked_at", "smartctl", "drives"} or
+    {"checked_at", "error"} when smartctl is missing or too old."""
+    base = {"checked_at": now_iso()}
+    version = smartctl_version()
+    if version is None:
+        return {**base, "error": "smartctl not installed"}
+    base["smartctl"] = "%d.%d" % version
+    if version < SMARTCTL_MIN:
+        return {**base, "error": "smartctl %d.%d is too old: drive health needs 7.0 or newer" % version}
+    _, scan = smartctl_json(["--scan-open"])
+    devices = (scan or {}).get("devices") if isinstance(scan, dict) else None
+    if not isinstance(devices, list):
+        return {**base, "error": "smartctl couldn't list the drives"}
+    drives = []
+    for d in devices[:DRIVES_MAX]:
+        name, kind = (d.get("name"), d.get("type")) if isinstance(d, dict) else (None, None)
+        if not isinstance(name, str) or not name.startswith("/dev/"):
+            continue
+        args = ["-a", name] if not isinstance(kind, str) else ["-a", "-d", kind, name]
+        status, data = smartctl_json(args)
+        drives.append(drive_summary(name, status, data, kind))
+    return {**base, "drives": drives}
+
+
+def write_drives(report: dict, path: str | None = None) -> None:
+    """DRIVES, written whole and readable by the agent."""
+    path = path or DRIVES
+    tmp = path + ".tmp"
+    with open(tmp, "w") as fh:
+        json.dump(report, fh)
+    os.chmod(tmp, 0o644)
+    os.replace(tmp, path)
+
+
+def write_result(result: dict) -> None:
+    """RESULTS/<id>.json, written whole (a temp file renamed into place)."""
+    os.makedirs(RESULTS, mode=0o755, exist_ok=True)
+    final = os.path.join(RESULTS, result["id"] + ".json")
+    tmp = final + ".tmp"
+    with open(tmp, "w") as fh:
+        json.dump(result, fh)
+    os.chmod(tmp, 0o644)
+    os.replace(tmp, final)
+
+
+def already_handled(job_id: str) -> bool:
+    """A request the agent sent again, for a job already run, is not run twice."""
+    return os.path.exists(os.path.join(RESULTS, job_id + ".json"))
+
+
+def handle(req: dict) -> dict:
+    base = {"id": req["id"], "action": req["action"], "started_at": now_iso()}
+    if req["action"] == "drives":               # a read, not a job: no result
+        write_drives(do_drives())
+        return {**base, "state": "done"}
+    write_result({**base, "state": "running"})
+    action = req["action"]
+    if action == "update":
+        outcome = do_update()
+    elif action in ("drain", "undrain"):
+        outcome = do_node("drain" if action == "drain" else "active", req.get("node"))
+    else:                                            # reboot
+        result = {**base, "state": "rebooting", "ok": True, "finished_at": now_iso()}
+        write_result(result)
+        os.sync()
+        run(["systemctl", "reboot"], deadline=60)
+        return result
+    result = {**base, **outcome, "state": "done" if outcome.get("ok") else "failed",
+              "finished_at": now_iso()}
+    write_result(result)
+    return result
+
+
+def prune_results(now: float | None = None) -> None:
+    now = time.time() if now is None else now
+    try:
+        names = os.listdir(RESULTS)
+    except OSError:
+        return
+    for name in names:
+        path = os.path.join(RESULTS, name)
+        try:
+            if now - os.lstat(path).st_mtime > RESULTS_KEPT_SECONDS:
+                os.unlink(path)
+        except OSError:
+            pass
+
+
+def pending_requests() -> list:
+    """Request files, oldest first."""
+    try:
+        names = [n for n in os.listdir(REQUESTS) if n.endswith(".json")]
+    except OSError:
+        return []
+    paths = [os.path.join(REQUESTS, n) for n in names]
+
+    def age(p):
+        try:
+            return os.lstat(p).st_mtime
+        except OSError:
+            return 0.0
+    return sorted(paths, key=age)
+
+
+def main() -> int:
+    prune_results()
+    for path in pending_requests():
+        req = read_request(path)
+        try:
+            os.unlink(path)                 # whatever it held: never re-read in a loop
+        except OSError:
+            pass
+        if req is None:
+            sys.stderr.write(f"[steward-maint] ignored {os.path.basename(path)}: not a valid request\n")
+            continue
+        if already_handled(req["id"]):
+            continue
+        result = handle(req)
+        sys.stderr.write(f"[steward-maint] {req['action']} {req['id']}: {result['state']}\n")
+        if result["state"] == "rebooting":
+            break                           # the rest waits for the next boot
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())

plugins/host_agent/files/steward-agent.service

@@ -0,0 +1,26 @@
+# Served by Steward at /plugins/host_agent/files/steward-agent.service and
+# installed by install.sh and the host_agent playbooks: one copy of the unit.
+[Unit]
+Description=Steward host agent
+After=network-online.target
+Wants=network-online.target
+
+[Service]
+Type=simple
+User=steward-agent
+Environment=STEWARD_AGENT_CONFIG=/etc/steward-agent.conf
+ExecStart=/usr/bin/env python3 /usr/local/lib/steward-agent/agent.py
+Restart=on-failure
+RestartSec=5
+# Stopping waits for the agent's last post, bounded at 10 s.
+TimeoutStopSec=30
+NoNewPrivileges=true
+ProtectSystem=strict
+ProtectHome=true
+PrivateTmp=true
+ReadOnlyPaths=/proc /sys
+# The one place the agent writes: requests to the maintenance helper (#5766).
+ReadWritePaths=-/var/lib/steward-agent/requests
+
+[Install]
+WantedBy=multi-user.target

plugins/host_agent/files/steward-maint.path

@@ -0,0 +1,10 @@
+# Starts the maintenance helper when the host agent leaves a request (#5766).
+[Unit]
+Description=Steward maintenance requests from the host agent
+
+[Path]
+PathExistsGlob=/var/lib/steward-agent/requests/*.json
+Unit=steward-maint.service
+
+[Install]
+WantedBy=paths.target

plugins/host_agent/files/steward-maint.service

@@ -0,0 +1,11 @@
+# Runs as root, and does only what maint.py's fixed list allows: update the
+# host's packages, reboot it, drain / reactivate its swarm node (#5766), or
+# read its drives' health with smartctl (#5789).
+[Unit]
+Description=Steward maintenance helper
+
+[Service]
+Type=oneshot
+ExecStart=/usr/bin/env python3 /usr/local/lib/steward-agent/maint.py
+# A full upgrade on a slow mirror; past this systemd stops it.
+TimeoutStartSec=3600

plugins/host_agent/maintenance.py

@@ -0,0 +1,438 @@
+# plugins/host_agent/maintenance.py
+"""A host's maintenance job, followed to its end through any restart (#5767).
+
+Steward plans and the host's agent acts: a job is one host's update and, only
+when the update says one is needed, its reboot, with the host's swarm node
+drained first and made active again after. Each action is one step the agent
+hands to its root helper (maint.py, #5766). Everything lives on the job's row,
+so Steward restarting, or the host Steward runs on rebooting, loses nothing:
+the next post from the agent carries the step's result, and the `started`
+event a host sends after booting is what completes a reboot.
+
+The transitions here are model-free (a job is anything with the attributes),
+so they're unit-tested without a database; routes.ingest and the scheduler's
+tick apply them.
+
+States: queued → updating → [draining →] rebooting → [undraining →] done,
+or failed / timed_out. A job with no reboot needed goes updating → done.
+"""
+from __future__ import annotations
+
+from datetime import datetime, timedelta, timezone
+
+ACTIVE = ("queued", "updating", "draining", "rebooting", "undraining")
+FINISHED = ("done", "failed", "timed_out")
+# The helper action each handed state asks for.
+STEP_ACTION = {"updating": "update", "draining": "drain", "rebooting": "reboot",
+               "undraining": "undrain"}
+# How long a state may last before the job is given up. Queued counts from the
+# job's creation (the host may be silent); the rest from their hand-off.
+DEADLINES = {
+    "queued": timedelta(hours=1),
+    "updating": timedelta(minutes=90),     # the helper itself stops at 60
+    "draining": timedelta(minutes=10),
+    "rebooting": timedelta(minutes=30),
+    "undraining": timedelta(minutes=10),
+}
+# A host's clock and Steward's differ by seconds; a boot this far before the
+# reboot was handed out is still taken as that reboot.
+BOOT_SLACK = timedelta(seconds=60)
+DETAIL_MAX = 500
+
+
+def step_id(job, state: str) -> str:
+    """The helper request id for a job's step: unique per job and step, and in
+    the helper's id alphabet (letters, digits, dashes)."""
+    return f"{job.id}-{STEP_ACTION[state]}"
+
+
+def _enter(job, state: str, now: datetime) -> None:
+    job.state, job.state_at = state, now
+    job.step_id = step_id(job, state) if state in STEP_ACTION else None
+    job.step_seen = False               # a new step is handed out until its result arrives
+    if state in FINISHED:
+        job.finished_at = now
+
+
+def _finish(job, state: str, now: datetime, detail: str | None = None) -> None:
+    if detail:
+        job.detail = detail[:DETAIL_MAX]
+    _enter(job, state, now)
+
+
+def start(job, now: datetime) -> None:
+    """The host's agent is about to be handed the job: update first."""
+    if job.state == "queued":
+        _enter(job, "updating", now)
+
+
+def executor(job) -> str:
+    """The host whose agent carries out the current step: the job's own host,
+    except a drain or undrain, which a swarm manager does."""
+    if job.state in ("draining", "undraining") and job.drain_host_id:
+        return job.drain_host_id
+    return job.host_id
+
+
+def handoff(job) -> dict | None:
+    """What to hand the executing agent now, or None. Handed on every post
+    until a result for the step arrives; the agent and the helper both ignore
+    an id they've already taken, so a repeat is harmless."""
+    if job.state not in STEP_ACTION or job.step_seen:
+        return None
+    out = {"id": job.step_id, "action": STEP_ACTION[job.state]}
+    if job.state in ("draining", "undraining") and job.node_id:
+        out["node"] = job.node_id
+    return out
+
+
+REBOOT_POLICIES = ("if_required", "always", "never")
+
+
+def wants_reboot(job) -> bool:
+    """Reboot only when the update says one is needed (the operator's
+    default), always (a plan may ask), or never."""
+    if job.reboot_policy == "always":
+        return True
+    return bool(job.reboot_needed) and job.reboot_policy == "if_required"
+
+
+def _after_update(job, now: datetime) -> None:
+    if not wants_reboot(job):
+        _enter(job, "done", now)
+    elif job.node_id and job.drain_host_id:
+        _enter(job, "draining", now)
+    else:
+        _enter(job, "rebooting", now)
+
+
+def apply_result(job, event: dict, now: datetime) -> bool:
+    """Apply a `maintenance` event for this job's current step; True when the
+    job changed. A result for a step the job has moved past is ignored, so a
+    repeated or late report does nothing."""
+    if job.state not in STEP_ACTION or event.get("id") != job.step_id:
+        return False
+    state = event.get("state")
+    if state in ("running", "rebooting"):
+        if job.step_seen:
+            return False
+        job.step_seen = True
+        return True
+    if state not in ("done", "failed"):
+        return False
+    if state == "failed" or event.get("ok") is False:
+        what = STEP_ACTION[job.state]
+        detail = f"{what} failed: {event.get('detail') or 'no detail'}"
+        if job.state == "undraining":
+            detail += " (the node is still drained)"
+        _finish(job, "failed", now, detail)
+        return True
+    if job.state == "updating":
+        if isinstance(event.get("packages"), int):
+            job.packages = event["packages"]
+        job.reboot_needed = event.get("reboot_required") is True
+        _after_update(job, now)
+    elif job.state == "draining":
+        _enter(job, "rebooting", now)
+    elif job.state == "undraining":
+        _enter(job, "done", now)
+    else:                               # rebooting: the boot completes it, not this
+        return False
+    return True
+
+
+def apply_boot(job, boot_at: datetime | None, now: datetime) -> bool:
+    """A `started` event from the job's host. A boot after the reboot was
+    handed out is that reboot: the host is back."""
+    if job.state != "rebooting" or boot_at is None or job.state_at is None:
+        return False
+    if boot_at < job.state_at - BOOT_SLACK:
+        return False                    # the agent restarted; the host didn't
+    job.back_at = now
+    if job.node_id and job.drain_host_id:
+        _enter(job, "undraining", now)
+    else:
+        _enter(job, "done", now)
+    return True
+
+
+def check_deadline(job, now: datetime) -> bool:
+    """Give up on a job whose state has lasted past its deadline."""
+    limit = DEADLINES.get(job.state)
+    since = job.state_at or job.created_at
+    if limit is None or since is None or now - since <= limit:
+        return False
+    waiting = {"queued": "the host's agent didn't check in",
+               "updating": "the update didn't finish",
+               "draining": "the node wasn't drained",
+               "rebooting": "the host didn't come back",
+               "undraining": "the node wasn't made active again"}[job.state]
+    _finish(job, "timed_out", now, f"{waiting} within {int(limit.total_seconds() // 60)} min")
+    return True
+
+
+def window_end(job, now: datetime) -> datetime:
+    """When the job's maintenance window should end: the longest the rest of
+    the job could take, or now once it has finished."""
+    if job.state in FINISHED:
+        return now
+    left = sum((DEADLINES[s] for s in ACTIVE[ACTIVE.index(job.state):]), timedelta())
+    return now + left
+
+
+def utcnow() -> datetime:
+    return datetime.now(timezone.utc)
+
+
+# ── applied to the database: ingest, the tick, and starting a job ────────────
+
+async def _window(session, job, host_name: str, now: datetime) -> None:
+    """Keep the job's maintenance window over the host until the job ends: one
+    scoped to this host's agent metrics, so an alert the job causes waits.
+    A swarm node's job keeps a second over the swarm-wide figures (#5788):
+    draining and rebooting a node takes it down and can leave a service
+    pinned to it short, which is the job, not news."""
+    from steward.models.alerts import MaintenanceWindow
+    if job.window_id:
+        w = await session.get(MaintenanceWindow, job.window_id)
+        if w is not None:
+            w.end_at = max(w.start_at, window_end(job, now))
+    elif job.created_by is None or job.state in FINISHED:
+        return                          # a window needs an owner; none for a finished job
+    else:
+        w = MaintenanceWindow(name=f"Maintenance: {host_name}"[:128], start_at=now,
+                              end_at=window_end(job, now), scope_source="host_agent",
+                              scope_resource=host_name, created_by=job.created_by)
+        session.add(w)
+        await session.flush()
+        job.window_id = w.id
+    if job.node_id:
+        await _swarm_window(session, job, host_name, now)
+
+
+SWARM_SCOPE = ("docker", "swarm")       # the swarm-wide figures' source and resource
+
+
+async def _swarm_window(session, job, host_name: str, now: datetime) -> None:
+    """The swarm half of a swarm node's job window, found again by its name
+    and the job's start, so it needs no column of its own."""
+    from sqlalchemy import select
+    from steward.models.alerts import MaintenanceWindow
+    name = f"Maintenance: {host_name} (swarm)"[:128]
+    w = (await session.execute(select(MaintenanceWindow).where(
+        MaintenanceWindow.name == name, MaintenanceWindow.scope_source == SWARM_SCOPE[0],
+        MaintenanceWindow.scope_resource == SWARM_SCOPE[1],
+        MaintenanceWindow.start_at >= job.created_at))).scalars().first()
+    if w is not None:
+        w.end_at = max(w.start_at, window_end(job, now))
+    elif job.created_by is not None and job.state not in FINISHED:
+        session.add(MaintenanceWindow(name=name, start_at=now, end_at=window_end(job, now),
+                                      scope_source=SWARM_SCOPE[0], scope_resource=SWARM_SCOPE[1],
+                                      created_by=job.created_by))
+
+
+async def _host_name(session, host_id: str) -> str:
+    from steward.models.hosts import Host
+    host = await session.get(Host, host_id)
+    return host.name if host is not None else host_id
+
+
+async def apply_events(session, host_id: str, events, now: datetime) -> int:
+    """Apply one post's `maintenance` results and `started` boots to the jobs
+    they belong to; returns how many jobs changed. A result is matched by its
+    step id, so a drain reported by a manager reaches the job of the node it
+    drained."""
+    from sqlalchemy import select
+    from .downtime import _at
+    from .models import HostMaintenanceJob
+    if not isinstance(events, list):
+        return 0
+    changed = set()
+    for event in events:
+        if not isinstance(event, dict):
+            continue
+        if event.get("kind") == "maintenance" and isinstance(event.get("id"), str):
+            job = (await session.execute(select(HostMaintenanceJob).where(
+                HostMaintenanceJob.step_id == event["id"],
+                HostMaintenanceJob.state.in_(ACTIVE)))).scalar_one_or_none()
+            if job is not None and apply_result(job, event, now):
+                changed.add(job)
+        elif event.get("kind") == "started":
+            boot_at = _at(event.get("boot_at"))
+            for job in (await session.execute(select(HostMaintenanceJob).where(
+                    HostMaintenanceJob.host_id == host_id,
+                    HostMaintenanceJob.state == "rebooting"))).scalars():
+                if apply_boot(job, boot_at, now):
+                    changed.add(job)
+    for job in changed:
+        await _window(session, job, await _host_name(session, job.host_id), now)
+    return len(changed)
+
+
+async def handoffs(session, host_id: str, host_name: str, now: datetime) -> list[dict]:
+    """The steps this host's agent should carry out now. A queued job for the
+    host starts here, when its agent is known to be listening."""
+    from sqlalchemy import or_, select
+    from .models import HostMaintenanceJob
+    jobs = (await session.execute(select(HostMaintenanceJob).where(
+        HostMaintenanceJob.state.in_(ACTIVE),
+        or_(HostMaintenanceJob.host_id == host_id,
+            HostMaintenanceJob.drain_host_id == host_id)
+    ).order_by(HostMaintenanceJob.created_at))).scalars().all()
+    out = []
+    for job in jobs:
+        if job.state == "queued" and job.host_id == host_id:
+            start(job, now)
+            await _window(session, job, host_name, now)
+        if executor(job) == host_id:
+            step = handoff(job)
+            if step is not None:
+                out.append(step)
+    return out
+
+
+async def tick(session, now: datetime) -> int:
+    """Give up on jobs past their deadline; returns how many."""
+    from sqlalchemy import select
+    from .models import HostMaintenanceJob
+    gave_up = 0
+    for job in (await session.execute(select(HostMaintenanceJob).where(
+            HostMaintenanceJob.state.in_(ACTIVE)))).scalars():
+        if check_deadline(job, now):
+            gave_up += 1
+            await _window(session, job, await _host_name(session, job.host_id), now)
+    return gave_up
+
+
+class CannotStart(Exception):
+    """Why a host can't take a maintenance job now, in words for the page."""
+
+
+def pick_drainer(node_hostname_to_id: dict, managers: dict, host_name: str,
+                 host_id: str, helpers: set, availability: str = "active") -> tuple:
+    """(node_id, drain_host_id) for a host in a swarm, (None, None) for one
+    that isn't. `managers` maps a manager's host id to its swarm node id;
+    `helpers` is the hosts whose agent has the maintenance helper. The host
+    drains itself when it is a manager, else another manager does. A node the
+    operator already set to drain or pause (a manager kept free of work) is
+    left as it is: no drain before, and no making it active after."""
+    node_id = node_hostname_to_id.get(host_name.lower())
+    if node_id is None:
+        return None, None
+    if availability != "active":
+        return node_id, None
+    if host_id in managers and host_id in helpers:
+        return node_id, host_id
+    for mgr in sorted(managers):
+        if mgr != host_id and mgr in helpers:
+            return node_id, mgr
+    raise CannotStart("it is a swarm node and no swarm manager has the maintenance "
+                      "helper to drain it: run Update agent on a manager")
+
+
+async def plan_job(session, host, reg, reboot_policy: str = "if_required") -> tuple:
+    """(node_id, drain_host_id) for a job on `host`, or CannotStart saying why
+    it can't have one now. The panel asks this before offering the button."""
+    from sqlalchemy import select
+    from .models import HostAgentRegistration, HostMaintenanceJob
+    if reg is None or not reg.maint_helper:
+        raise CannotStart("its agent has no maintenance helper: run Update agent first")
+    if reboot_policy not in REBOOT_POLICIES:
+        raise CannotStart(f"unknown reboot policy {reboot_policy!r}")
+    busy = (await session.execute(select(HostMaintenanceJob.id).where(
+        HostMaintenanceJob.host_id == host.id,
+        HostMaintenanceJob.state.in_(ACTIVE)))).first()
+    if busy is not None:
+        raise CannotStart("a maintenance job is already running on it")
+    if reboot_policy == "never":
+        return None, None
+    nodes, managers, availability = await _swarm(session)
+    helpers = set((await session.execute(select(HostAgentRegistration.host_id).where(
+        HostAgentRegistration.maint_helper.is_(True)))).scalars())
+    node_id = nodes.get(host.name.lower())
+    return pick_drainer(nodes, managers, host.name, host.id, helpers,
+                        availability.get(node_id, "active"))
+
+
+async def create_job(session, host, reg, user_id: str | None,
+                     reboot_policy: str = "if_required", now: datetime | None = None,
+                     run_id: str | None = None):
+    """Queue a maintenance job for `host`; it starts with the agent's next post."""
+    from .models import HostMaintenanceJob
+    now = now or utcnow()
+    node_id, drain_host_id = await plan_job(session, host, reg, reboot_policy)
+    job = HostMaintenanceJob(host_id=host.id, state="queued", state_at=now,
+                             reboot_policy=reboot_policy, node_id=node_id,
+                             drain_host_id=drain_host_id, created_by=user_id, created_at=now,
+                             run_id=run_id)
+    session.add(job)
+    await session.flush()
+    return job
+
+
+async def latest_job(session, host_id: str, now: datetime, days: int = 366):
+    """The host's running job, else its last one (within a year), for the panel."""
+    from sqlalchemy import select
+    from .models import HostMaintenanceJob
+    return (await session.execute(select(HostMaintenanceJob).where(
+        HostMaintenanceJob.host_id == host_id,
+        HostMaintenanceJob.created_at >= now - timedelta(days=days),
+    ).order_by(HostMaintenanceJob.state.in_(ACTIVE).desc(),
+               HostMaintenanceJob.created_at.desc()).limit(1))).scalar_one_or_none()
+
+
+LABELS = {
+    "queued": "Waiting for the agent",
+    "updating": "Updating",
+    "draining": "Draining its swarm node",
+    "rebooting": "Rebooting",
+    "undraining": "Making its swarm node active again",
+    "done": "Done",
+    "failed": "Failed",
+    "timed_out": "Timed out",
+}
+
+
+def summary(job) -> str:
+    """One line on what the job did, for the panel."""
+    if job.state not in FINISHED:
+        return LABELS[job.state]
+    parts = []
+    if job.packages is not None:
+        parts.append(f"{job.packages} package{'s' if job.packages != 1 else ''} updated")
+    if job.back_at is not None:
+        parts.append("rebooted")
+    elif job.reboot_needed and job.reboot_policy == "never":
+        parts.append("reboot needed, not allowed by this job")
+    elif job.state == "done" and job.reboot_needed is False:
+        parts.append("no reboot needed")
+    head = LABELS[job.state]
+    return f"{head}: {', '.join(parts)}" if parts else head
+
+
+def docker_models():
+    """The loaded Docker plugin's models module, or None when it isn't loaded."""
+    from steward.core.plugin_manager import loaded_plugin_submodule
+    return loaded_plugin_submodule("docker", "models")
+
+
+async def _swarm(session) -> tuple:
+    """({node hostname: node id}, {manager host id: its node id},
+    {node id: availability}) from what the Docker plugin last stored; empty
+    when it isn't installed. A manager is a host whose agent reported the
+    swarm's nodes."""
+    models = docker_models()
+    if models is None:
+        return {}, {}, {}
+    DockerSwarmNode = models.DockerSwarmNode
+    from sqlalchemy import select
+    from steward.models.hosts import Host
+    rows = (await session.execute(select(DockerSwarmNode))).scalars().all()
+    nodes = {(n.hostname or "").lower(): n.node_id for n in rows if n.hostname}
+    reporters = {n.host_id for n in rows}
+    names = dict((await session.execute(select(Host.id, Host.name).where(
+        Host.id.in_(reporters)))).all()) if reporters else {}
+    managers = {hid: nodes.get(name.lower()) for hid, name in names.items()}
+    availability = {n.node_id: (n.availability or "active")
+                    for n in sorted(rows, key=lambda n: n.updated_at)}
+    return nodes, managers, availability

plugins/host_agent/mcp_tools.py

@@ -0,0 +1,100 @@
+"""The host_agent plugin's MCP tools and its get_overview findings."""
+from __future__ import annotations
+
+from datetime import timedelta
+
+from sqlalchemy import select
+
+from steward.hosts import service as host_svc
+from steward.mcp._context import mcp_app
+from steward.mcp.tools._shape import age_s, db_session, iso, now_utc
+from steward.mcp.tools.overview import finding
+from steward.models.hosts import Host
+from steward.models.users import UserRole
+
+from .models import HostAgentRegistration
+
+# A registration this old that has never checked in is a failed install, not one in progress.
+PENDING_GRACE = timedelta(minutes=30)
+
+
+def agent_state(reg, now, stale_after_s: int) -> str:
+    """The agent's state: reporting, stale (stopped checking in) or pending (never has)."""
+    if reg.last_seen_at is None:
+        return "pending"
+    return "reporting" if (now - reg.last_seen_at).total_seconds() <= stale_after_s else "stale"
+
+
+async def _agents(db):
+    return (await db.execute(
+        select(HostAgentRegistration, Host)
+        .join(Host, Host.id == HostAgentRegistration.host_id)
+        .order_by(Host.name)
+    )).all()
+
+
+def _updates(reg) -> dict | None:
+    """A registration's pending updates and reboot, as list_agents says them (#5765)."""
+    if reg.updates_checked_at is None and reg.reboot_required is None:
+        return None
+    return {
+        "manager": reg.updates_manager, "pending": reg.updates_pending,
+        "security": reg.updates_security, "checked_at": iso(reg.updates_checked_at),
+        "error": reg.updates_error, "reboot_required": reg.reboot_required,
+        "reboot_required_since": iso(reg.reboot_required_since),
+        "reboot_packages": reg.reboot_packages.split(", ") if reg.reboot_packages else [],
+    }
+
+
+async def list_agents() -> dict:
+    """Every host agent Steward has registered, with its version and whether it is reporting.
+
+    `state` is "reporting", "stale" (it checked in before but its last
+    check-in is older than the stale window, `stale_after_s`) or "pending"
+    (registered, never checked in, usually an install that is still running
+    or that failed). Each row carries the agent version, kernel, distro, arch
+    and the IP the agent reports. get_host shows an agent's latest metrics.
+
+    `updates` is what the host's package manager has waiting, from the
+    agent's hourly check (agent ≥ 1.16.0): manager, pending, security (null
+    when the manager can't say), checked_at, error, and reboot_required with
+    since when and the packages that asked for it. null before the first check.
+    """
+    now = now_utc()
+    stale_after = host_svc.agent_stale_after_seconds(mcp_app().config)
+    async with db_session() as db:
+        rows = await _agents(db)
+    agents = [
+        {
+            "host_id": h.id, "host_name": h.name,
+            "state": agent_state(reg, now, stale_after),
+            "last_seen_at": iso(reg.last_seen_at), "last_seen_age_s": age_s(reg.last_seen_at, now),
+            "agent_version": reg.agent_version, "kernel": reg.kernel, "distro": reg.distro,
+            "arch": reg.arch, "reported_ip": reg.host_ip,
+            "registered_at": iso(reg.created_at),
+            "updates": _updates(reg),
+        }
+        for reg, h in rows
+    ]
+    counts = {s: sum(1 for a in agents if a["state"] == s) for s in ("reporting", "stale", "pending")}
+    return {"stale_after_s": stale_after, "counts": counts, "agents": agents, "total": len(agents)}
+
+
+async def mcp_overview(db) -> list[dict]:
+    """Agents registered more than PENDING_GRACE ago that have never checked in.
+
+    Agents that stopped reporting are already flagged by the core overview
+    from their metrics, so they are not repeated here.
+    """
+    now = now_utc()
+    return [
+        finding("warning", "agent_never_reported", h.name,
+                "host agent was registered but has never checked in; the install may have failed",
+                since=reg.created_at, ids={"host_id": h.id}, next="list_agents()")
+        for reg, h in await _agents(db)
+        if reg.last_seen_at is None and now - reg.created_at > PENDING_GRACE
+    ]
+
+
+def register_tools(register) -> None:
+    register(list_agents, role=UserRole.viewer, read_only=True)

plugins/host_agent/metrics_query.py

@@ -9,7 +9,7 @@ from __future__ import annotations
 
 from datetime import datetime, timedelta, timezone
 
-from sqlalchemy import func, or_, select
+from sqlalchemy import func, select
 
 from steward.core.time_range import bucket_seconds
 from steward.models.metrics import PluginMetric, PluginMetricHourly
@@ -27,30 +27,14 @@ HISTORY_METRICS = (
 
 async def _latest_metrics_for_host(session, host_name: str) -> dict[str, dict[str, float]]:
     """{resource_name: {metric: value}} — latest sample per (resource, metric) for
-    a host + its sub-resources.
-
-    DISTINCT ON picks the newest row per group in one index-ordered pass over
-    ix_plugin_metrics_module_resource_metric_recorded, instead of a GROUP-BY-max
-    subquery self-joined back to the table (two passes over the whole history).
+    a host + its sub-resources, from plugin_metric_latest (#5208). A host whose
+    agent has gone quiet keeps its last values until they age out of the raw
+    window; the caller decides staleness from the registration's last_seen_at.
     """
-    rows = (await session.execute(
-        select(PluginMetric)
-        .where(
-            PluginMetric.source_module == SOURCE_MODULE,
-            or_(
-                PluginMetric.resource_name == host_name,
-                PluginMetric.resource_name.like(host_name + ":%"),
-            ),
-        )
-        .distinct(PluginMetric.resource_name, PluginMetric.metric_name)
-        .order_by(
-            PluginMetric.resource_name,
-            PluginMetric.metric_name,
-            PluginMetric.recorded_at.desc(),
-        )
-    )).scalars().all()
+    from steward.core.metric_series import latest_values
+
     out: dict[str, dict[str, float]] = {}
-    for r in rows:
+    for r in await latest_values(session, SOURCE_MODULE, host_name, include_children=True):
         out.setdefault(r.resource_name, {})[r.metric_name] = r.value
     return out
 

plugins/host_agent/migrations/versions/host_agent_003_clock_offset.py

@@ -0,0 +1,24 @@
+# plugins/host_agent/migrations/versions/host_agent_003_clock_offset.py
+"""host_agent: how far each host's clock is from Steward's (milestone 490)
+
+Revision ID: host_agent_003_clock_offset
+"""
+from typing import Sequence, Union
+from alembic import op
+import sqlalchemy as sa
+
+revision: str = "host_agent_003_clock_offset"
+down_revision: Union[str, None] = "host_agent_002_host_ip"
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    op.add_column("host_agent_registrations", sa.Column("clock_offset_s", sa.Float(), nullable=True))
+    op.add_column("host_agent_registrations",
+                  sa.Column("clock_checked_at", sa.DateTime(timezone=True), nullable=True))
+
+
+def downgrade() -> None:
+    op.drop_column("host_agent_registrations", "clock_checked_at")
+    op.drop_column("host_agent_registrations", "clock_offset_s")

plugins/host_agent/migrations/versions/host_agent_004_downtime.py

@@ -0,0 +1,32 @@
+# plugins/host_agent/migrations/versions/host_agent_004_downtime.py
+"""host_agent: each host's downtime and what it was (#5756)
+
+Revision ID: host_agent_004_downtime
+"""
+from typing import Sequence, Union
+from alembic import op
+import sqlalchemy as sa
+
+revision: str = "host_agent_004_downtime"
+down_revision: Union[str, None] = "host_agent_003_clock_offset"
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    op.create_table(
+        "host_downtime",
+        sa.Column("id", sa.String(36), primary_key=True),
+        sa.Column("host_id", sa.String(36), sa.ForeignKey("hosts.id", ondelete="CASCADE"),
+                  nullable=False),
+        sa.Column("kind", sa.String(16), nullable=False),
+        sa.Column("started_at", sa.DateTime(timezone=True), nullable=False),
+        sa.Column("ended_at", sa.DateTime(timezone=True), nullable=True),
+        sa.Column("created_at", sa.DateTime(timezone=True), nullable=False),
+    )
+    op.create_index("ix_host_downtime_host_id", "host_downtime", ["host_id"])
+
+
+def downgrade() -> None:
+    op.drop_index("ix_host_downtime_host_id", table_name="host_downtime")
+    op.drop_table("host_downtime")

plugins/host_agent/migrations/versions/host_agent_005_updates.py

@@ -0,0 +1,34 @@
+# plugins/host_agent/migrations/versions/host_agent_005_updates.py
+"""host_agent: a host's pending updates and whether it needs a reboot (#5765)
+
+Revision ID: host_agent_005_updates
+"""
+from typing import Sequence, Union
+from alembic import op
+import sqlalchemy as sa
+
+revision: str = "host_agent_005_updates"
+down_revision: Union[str, None] = "host_agent_004_downtime"
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+_COLUMNS = (
+    ("updates_manager", sa.String(8)),
+    ("updates_pending", sa.Integer()),
+    ("updates_security", sa.Integer()),
+    ("updates_checked_at", sa.DateTime(timezone=True)),
+    ("updates_error", sa.String(200)),
+    ("reboot_required", sa.Boolean()),
+    ("reboot_required_since", sa.DateTime(timezone=True)),
+    ("reboot_packages", sa.Text()),
+)
+
+
+def upgrade() -> None:
+    for name, kind in _COLUMNS:
+        op.add_column("host_agent_registrations", sa.Column(name, kind, nullable=True))
+
+
+def downgrade() -> None:
+    for name, _kind in reversed(_COLUMNS):
+        op.drop_column("host_agent_registrations", name)

plugins/host_agent/migrations/versions/host_agent_006_maint_helper.py

@@ -0,0 +1,21 @@
+# plugins/host_agent/migrations/versions/host_agent_006_maint_helper.py
+"""host_agent: whether a host has the maintenance helper (#5766)
+
+Revision ID: host_agent_006_maint_helper
+"""
+from typing import Sequence, Union
+from alembic import op
+import sqlalchemy as sa
+
+revision: str = "host_agent_006_maint_helper"
+down_revision: Union[str, None] = "host_agent_005_updates"
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    op.add_column("host_agent_registrations", sa.Column("maint_helper", sa.Boolean(), nullable=True))
+
+
+def downgrade() -> None:
+    op.drop_column("host_agent_registrations", "maint_helper")

plugins/host_agent/migrations/versions/host_agent_007_maint_jobs.py

@@ -0,0 +1,47 @@
+"""host_agent: a host's maintenance job, followed through its reboot (#5767)
+
+Revision ID: host_agent_007_maint_jobs
+"""
+from typing import Sequence, Union
+from alembic import op
+import sqlalchemy as sa
+
+revision: str = "host_agent_007_maint_jobs"
+down_revision: Union[str, None] = "host_agent_006_maint_helper"
+branch_labels: Union[str, Sequence[str], None] = None
+# maintenance_windows (core 0010) must exist before the window_id key.
+depends_on: Union[str, Sequence[str], None] = ("0010_maintenance_windows",)
+
+
+def upgrade() -> None:
+    op.create_table(
+        "host_maintenance_jobs",
+        sa.Column("id", sa.String(36), primary_key=True),
+        sa.Column("host_id", sa.String(36), sa.ForeignKey("hosts.id", ondelete="CASCADE"),
+                  nullable=False),
+        sa.Column("state", sa.String(16), nullable=False),
+        sa.Column("state_at", sa.DateTime(timezone=True), nullable=True),
+        sa.Column("reboot_policy", sa.String(16), nullable=False),
+        sa.Column("step_id", sa.String(64), nullable=True),
+        sa.Column("step_seen", sa.Boolean(), nullable=False, server_default=sa.false()),
+        sa.Column("node_id", sa.String(64), nullable=True),
+        sa.Column("drain_host_id", sa.String(36), sa.ForeignKey("hosts.id", ondelete="SET NULL"),
+                  nullable=True),
+        sa.Column("packages", sa.Integer(), nullable=True),
+        sa.Column("reboot_needed", sa.Boolean(), nullable=True),
+        sa.Column("back_at", sa.DateTime(timezone=True), nullable=True),
+        sa.Column("detail", sa.Text(), nullable=True),
+        sa.Column("window_id", sa.String(36),
+                  sa.ForeignKey("maintenance_windows.id", ondelete="SET NULL"), nullable=True),
+        sa.Column("created_by", sa.String(36), sa.ForeignKey("users.id", ondelete="SET NULL"),
+                  nullable=True),
+        sa.Column("created_at", sa.DateTime(timezone=True), nullable=False),
+        sa.Column("finished_at", sa.DateTime(timezone=True), nullable=True),
+    )
+    op.create_index("ix_host_maintenance_jobs_host_id", "host_maintenance_jobs", ["host_id"])
+    op.create_index("ix_host_maintenance_jobs_state", "host_maintenance_jobs", ["state"])
+    op.create_index("ix_host_maintenance_jobs_step_id", "host_maintenance_jobs", ["step_id"])
+
+
+def downgrade() -> None:
+    op.drop_table("host_maintenance_jobs")

plugins/host_agent/migrations/versions/host_agent_008_maint_plans.py

@@ -0,0 +1,61 @@
+"""host_agent: a group's maintenance cadence and its runs (#5768)
+
+Revision ID: host_agent_008_maint_plans
+"""
+from typing import Sequence, Union
+from alembic import op
+import sqlalchemy as sa
+
+revision: str = "host_agent_008_maint_plans"
+down_revision: Union[str, None] = "host_agent_007_maint_jobs"
+branch_labels: Union[str, Sequence[str], None] = None
+# ansible_groups (core 0016) must exist before the group_id key.
+depends_on: Union[str, Sequence[str], None] = ("0016_ansible_inventory_groups",)
+
+
+def upgrade() -> None:
+    op.create_table(
+        "host_maintenance_plans",
+        sa.Column("id", sa.String(36), primary_key=True),
+        sa.Column("name", sa.String(128), nullable=False),
+        sa.Column("group_id", sa.String(36), sa.ForeignKey("ansible_groups.id", ondelete="CASCADE"),
+                  nullable=False),
+        sa.Column("weekday", sa.Integer(), nullable=False),
+        sa.Column("at_time", sa.String(5), nullable=False),
+        sa.Column("time_zone", sa.String(64), nullable=False, server_default="UTC"),
+        sa.Column("window_minutes", sa.Integer(), nullable=False, server_default="240"),
+        sa.Column("reboot_policy", sa.String(16), nullable=False, server_default="if_required"),
+        sa.Column("enabled", sa.Boolean(), nullable=False, server_default=sa.true()),
+        sa.Column("last_slot_at", sa.DateTime(timezone=True), nullable=True),
+        sa.Column("created_by", sa.String(36), sa.ForeignKey("users.id", ondelete="SET NULL"),
+                  nullable=True),
+        sa.Column("created_at", sa.DateTime(timezone=True), nullable=False),
+    )
+    op.create_table(
+        "host_maintenance_runs",
+        sa.Column("id", sa.String(36), primary_key=True),
+        sa.Column("plan_id", sa.String(36),
+                  sa.ForeignKey("host_maintenance_plans.id", ondelete="CASCADE"), nullable=False),
+        sa.Column("state", sa.String(16), nullable=False),
+        sa.Column("host_ids", sa.Text(), nullable=False, server_default="[]"),
+        sa.Column("position", sa.Integer(), nullable=False, server_default="0"),
+        sa.Column("skipped", sa.Text(), nullable=False, server_default="[]"),
+        sa.Column("settle_since", sa.DateTime(timezone=True), nullable=True),
+        sa.Column("detail", sa.Text(), nullable=True),
+        sa.Column("manual", sa.Boolean(), nullable=False, server_default=sa.false()),
+        sa.Column("started_at", sa.DateTime(timezone=True), nullable=False),
+        sa.Column("finished_at", sa.DateTime(timezone=True), nullable=True),
+    )
+    op.create_index("ix_host_maintenance_runs_plan_id", "host_maintenance_runs", ["plan_id"])
+    op.create_index("ix_host_maintenance_runs_state", "host_maintenance_runs", ["state"])
+    op.add_column("host_maintenance_jobs", sa.Column(
+        "run_id", sa.String(36), sa.ForeignKey("host_maintenance_runs.id", ondelete="SET NULL"),
+        nullable=True))
+    op.create_index("ix_host_maintenance_jobs_run_id", "host_maintenance_jobs", ["run_id"])
+
+
+def downgrade() -> None:
+    op.drop_index("ix_host_maintenance_jobs_run_id", "host_maintenance_jobs")
+    op.drop_column("host_maintenance_jobs", "run_id")
+    op.drop_table("host_maintenance_runs")
+    op.drop_table("host_maintenance_plans")

plugins/host_agent/migrations/versions/host_agent_009_drive_health.py

@@ -0,0 +1,24 @@
+# plugins/host_agent/migrations/versions/host_agent_009_drive_health.py
+"""host_agent: each host's newest drive-health report (#5789)
+
+Revision ID: host_agent_009_drive_health
+"""
+from typing import Sequence, Union
+from alembic import op
+import sqlalchemy as sa
+
+revision: str = "host_agent_009_drive_health"
+down_revision: Union[str, None] = "host_agent_008_maint_plans"
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    op.add_column("host_agent_registrations", sa.Column("drive_health", sa.JSON(), nullable=True))
+    op.add_column("host_agent_registrations",
+                  sa.Column("drive_health_at", sa.DateTime(timezone=True), nullable=True))
+
+
+def downgrade() -> None:
+    op.drop_column("host_agent_registrations", "drive_health_at")
+    op.drop_column("host_agent_registrations", "drive_health")

plugins/host_agent/models.py

@@ -2,7 +2,7 @@
 from __future__ import annotations
 import uuid
 from datetime import datetime, timezone
-from sqlalchemy import Column, String, DateTime, ForeignKey
+from sqlalchemy import JSON, Boolean, Column, String, DateTime, Float, ForeignKey, Integer, Text
 from steward.models.base import Base
 
 
@@ -29,6 +29,134 @@ class HostAgentRegistration(Base):
     # Agent-reported primary IP. 45 chars = max textual IPv6 (no zone suffix).
     host_ip = Column(String(45), nullable=True)
     last_seen_at = Column(DateTime(timezone=True), nullable=True)
+    # How far this host's clock is from Steward's, in seconds, positive when
+    # the host is ahead, measured from the agent's sent_at on the newest post;
+    # and when that was. None until an agent that stamps its posts reports.
+    clock_offset_s = Column(Float, nullable=True)
+    clock_checked_at = Column(DateTime(timezone=True), nullable=True)
+    # What the package manager has waiting, from the agent's hourly check
+    # (agent ≥ 1.16.0, #5765). updates_security is None when the manager
+    # can't say; updates_error is why the last check failed.
+    updates_manager = Column(String(8), nullable=True)
+    updates_pending = Column(Integer, nullable=True)
+    updates_security = Column(Integer, nullable=True)
+    updates_checked_at = Column(DateTime(timezone=True), nullable=True)
+    updates_error = Column(String(200), nullable=True)
+    # Whether the host needs a reboot (None: it can't say), since when, and
+    # the packages that asked for it, comma-separated.
+    reboot_required = Column(Boolean, nullable=True)
+    reboot_required_since = Column(DateTime(timezone=True), nullable=True)
+    reboot_packages = Column(Text, nullable=True)
+    # Whether the host has the root maintenance helper (steward-maint.path and
+    # its request directory) that updates and reboots go through (#5766); None
+    # until an agent that says reports.
+    maint_helper = Column(Boolean, nullable=True)
+    # The host's newest drive-health report, as the root helper read it with
+    # smartctl (agent ≥ 1.20.0, #5789): {"checked_at", "smartctl", "drives"} or
+    # {"error"}. drives.py judges it; the figures are also metrics.
+    drive_health = Column(JSON, nullable=True)
+    drive_health_at = Column(DateTime(timezone=True), nullable=True)
     created_at = Column(DateTime(timezone=True), nullable=False, default=_utcnow)
     updated_at = Column(DateTime(timezone=True), nullable=False,
                         default=_utcnow, onupdate=_utcnow)
+
+
+class HostDowntime(Base):
+    """A span a host's agent was away, and what the agent said it was (#5756).
+
+    kind is "reboot" (the host went down and came back), "agent_stop" (the
+    agent stopped; the host stayed up), "unclean" (the host came back from a
+    boot nobody announced: power loss or a crash) or "lost_contact" (the agent
+    ran but couldn't reach Steward). ended_at is None while it is going on.
+    """
+    __tablename__ = "host_downtime"
+
+    id = Column(String(36), primary_key=True, default=_uuid)
+    host_id = Column(String(36), ForeignKey("hosts.id", ondelete="CASCADE"),
+                     nullable=False, index=True)
+    kind = Column(String(16), nullable=False)
+    started_at = Column(DateTime(timezone=True), nullable=False)
+    ended_at = Column(DateTime(timezone=True), nullable=True)
+    created_at = Column(DateTime(timezone=True), nullable=False, default=_utcnow)
+
+
+class HostMaintenanceJob(Base):
+    """One host's update, and its reboot when the update needs one (#5767).
+
+    Steward keeps the job here and the host's agent carries out each step
+    through its root helper; maintenance.py has the states and transitions.
+    step_id is the helper request id of the current step, and step_seen says a
+    result for it has arrived, so it stops being handed out. node_id and
+    drain_host_id are set when the host is a swarm node: the node to drain and
+    the manager whose agent drains it.
+    """
+    __tablename__ = "host_maintenance_jobs"
+
+    id = Column(String(36), primary_key=True, default=_uuid)
+    host_id = Column(String(36), ForeignKey("hosts.id", ondelete="CASCADE"),
+                     nullable=False, index=True)
+    state = Column(String(16), nullable=False, default="queued", index=True)
+    state_at = Column(DateTime(timezone=True), nullable=True)
+    reboot_policy = Column(String(16), nullable=False, default="if_required")
+    step_id = Column(String(64), nullable=True, index=True)
+    step_seen = Column(Boolean, nullable=False, default=False)
+    node_id = Column(String(64), nullable=True)
+    drain_host_id = Column(String(36), ForeignKey("hosts.id", ondelete="SET NULL"), nullable=True)
+    packages = Column(Integer, nullable=True)
+    reboot_needed = Column(Boolean, nullable=True)
+    back_at = Column(DateTime(timezone=True), nullable=True)
+    detail = Column(Text, nullable=True)
+    window_id = Column(String(36), ForeignKey("maintenance_windows.id", ondelete="SET NULL"),
+                       nullable=True)
+    created_by = Column(String(36), ForeignKey("users.id", ondelete="SET NULL"), nullable=True)
+    created_at = Column(DateTime(timezone=True), nullable=False, default=_utcnow)
+    finished_at = Column(DateTime(timezone=True), nullable=True)
+    # The cadence run this job belongs to (#5768); None for Update now.
+    run_id = Column(String(36), ForeignKey("host_maintenance_runs.id", ondelete="SET NULL"),
+                    nullable=True, index=True)
+
+
+class HostMaintenancePlan(Base):
+    """A group's maintenance cadence (#5768): every week on `weekday` at
+    `at_time` in `time_zone`, its hosts are updated one at a time, each
+    rebooting per `reboot_policy`. No host is started after `window_minutes`
+    from the run's start. last_slot_at is the last scheduled slot that fired;
+    Run now leaves it alone, so the schedule keeps its time."""
+    __tablename__ = "host_maintenance_plans"
+
+    id = Column(String(36), primary_key=True, default=_uuid)
+    name = Column(String(128), nullable=False)
+    group_id = Column(String(36), ForeignKey("ansible_groups.id", ondelete="CASCADE"),
+                      nullable=False)
+    weekday = Column(Integer, nullable=False)              # 0 = Monday
+    at_time = Column(String(5), nullable=False)            # "HH:MM"
+    time_zone = Column(String(64), nullable=False, default="UTC")
+    window_minutes = Column(Integer, nullable=False, default=240)
+    reboot_policy = Column(String(16), nullable=False, default="if_required")
+    enabled = Column(Boolean, nullable=False, default=True)
+    last_slot_at = Column(DateTime(timezone=True), nullable=True)
+    created_by = Column(String(36), ForeignKey("users.id", ondelete="SET NULL"), nullable=True)
+    created_at = Column(DateTime(timezone=True), nullable=False, default=_utcnow)
+
+
+class HostMaintenanceRun(Base):
+    """One run of a plan: its hosts in order, one at a time (#5768).
+
+    host_ids is the order fixed at the start (JSON list); position is the host
+    being worked on; skipped is [{"host_id", "reason"}] for hosts the run
+    passed over. settle_since is when the current host's job finished and the
+    run began waiting for the host to be back in service."""
+    __tablename__ = "host_maintenance_runs"
+
+    id = Column(String(36), primary_key=True, default=_uuid)
+    plan_id = Column(String(36), ForeignKey("host_maintenance_plans.id", ondelete="CASCADE"),
+                     nullable=False, index=True)
+    state = Column(String(16), nullable=False, default="running", index=True)
+    host_ids = Column(Text, nullable=False, default="[]")
+    position = Column(Integer, nullable=False, default=0)
+    skipped = Column(Text, nullable=False, default="[]")
+    settle_since = Column(DateTime(timezone=True), nullable=True)
+    detail = Column(Text, nullable=True)
+    manual = Column(Boolean, nullable=False, default=False)
+    started_at = Column(DateTime(timezone=True), nullable=False, default=_utcnow)
+    finished_at = Column(DateTime(timezone=True), nullable=True)

plugins/host_agent/plugin.yaml

@@ -16,4 +16,7 @@ tags:
 
 config:
   stale_after_seconds: 180
+  # How long a reboot or restart the agent announced explains its silence
+  # before "agent offline" fires (agent >= 1.14.0, #5756).
+  reboot_grace_seconds: 600
   default_interval_seconds: 30

plugins/host_agent/scheduler.py

@@ -1,78 +1,177 @@
 # plugins/host_agent/scheduler.py
 from __future__ import annotations
 import logging
-from datetime import datetime, timedelta, timezone
+from datetime import datetime, timezone
 from typing import Iterable
 
 from sqlalchemy import select
 
+from steward.core.alerts import MetricBatch
 from steward.core.scheduler import ScheduledTask
+from steward.core.watch import watch
+from steward.hosts.service import agent_stale_after_seconds
 from steward.models.hosts import Host
+from steward.models.metrics import PluginMetricLatest
+from .downtime import announced_stops, restart_grace_seconds
 from .models import HostAgentRegistration
 
 logger = logging.getLogger(__name__)
 
 
-def _filter_stale(
-    regs: Iterable,
-    *,
-    now: datetime,
-    stale_after_seconds: int,
-) -> list:
-    """Pure staleness filter: returns the subset with last_seen_at strictly
-    older than (now - stale_after_seconds). Rows with last_seen_at=None are
-    never stale (they are unregistered-in-practice)."""
-    cutoff = now - timedelta(seconds=stale_after_seconds)
-    return [
-        r for r in regs
-        if r.last_seen_at is not None and r.last_seen_at < cutoff
-    ]
+def arrivals(app) -> dict[str, datetime]:
+    """{host id: when its agent's newest post reached Steward}, in memory (#5514).
+
+    last_seen_at is written in the ingest's own transaction, so when Steward
+    can't store a post (its storage stalls and the request is dropped) the
+    agent's report is lost with it, and an agent posting every few seconds read
+    as offline. A post that arrived is the agent reaching Steward; storing it
+    is Steward's part. Set before any write, so it holds however slow they are.
+    """
+    got = getattr(app, "_agent_arrivals", None)
+    if got is None:
+        got = app._agent_arrivals = {}
+    return got
 
 
-async def find_stale_registrations(app, stale_after_seconds: int = 180) -> list[dict]:
+def seconds_since_seen(regs: Iterable, host_names: dict[str, str],
+                       now: datetime, watching_since: datetime | None = None,
+                       arrived: dict[str, datetime] | None = None,
+                       already_silent: set[str] | frozenset = frozenset(),
+                       explained_until: dict[str, datetime] | None = None,
+                       ) -> list[tuple[str, float]]:
+    """(host name, seconds since its agent last reported) per registration.
+
+    An agent that has never reported counts from when it was registered. A
+    registration whose host is gone is skipped. Nothing counts from before
+    `watching_since`, when Steward last started watching without a break: an
+    agent can't report while Steward is down, so that time is not its silence
+    (#5441). An agent that is really gone reads as offline once it has been
+    quiet that long while Steward watched. `arrived` (arrivals(app)) counts a
+    post Steward received but couldn't store as the agent seen (#5514).
+
+    `already_silent` names hosts whose last recorded figure was already past
+    the offline window: their agent was gone before Steward stopped, so their
+    whole silence counts. Counting them from `watching_since` read them as
+    just seen, which resolved their alert on every restart and fired it again
+    three minutes later (#5719).
+
+    `explained_until` ({host id: time}) is a stop the agent announced, a reboot
+    or its own restart, plus its grace (#5756): until then the agent reads as
+    just seen. A host still silent past it reads its whole silence, so agent
+    offline fires as the grace runs out, for silence nobody explained.
+    """
+    out = []
+    for r in regs:
+        name = host_names.get(r.host_id)
+        seen = max((t for t in (r.last_seen_at or r.created_at, (arrived or {}).get(r.host_id))
+                    if t is not None), default=None)
+        if name is None or seen is None:
+            continue
+        if watching_since is not None and name not in already_silent:
+            seen = max(seen, watching_since)
+        if r.host_id in (explained_until or {}):
+            seen = max(seen, explained_until[r.host_id])
+        out.append((name, max(0.0, (now - seen).total_seconds())))
+    return out
+
+
+async def record_seconds_since_seen(app) -> None:
+    """Record host_agent/<host>/seconds_since_seen for every registered agent,
+    so "agent offline" is an ordinary alert rule (seconds_since_seen > 180)."""
+    now = datetime.now(timezone.utc)
     async with app.db_sessionmaker() as session:
-        all_regs = (await session.execute(
-            select(HostAgentRegistration)
-        )).scalars().all()
-        stale_rows = _filter_stale(
-            all_regs,
-            now=datetime.now(timezone.utc),
-            stale_after_seconds=stale_after_seconds,
-        )
-        if not stale_rows:
-            return []
-        hosts = {
-            h.id: h for h in (await session.execute(
-                select(Host).where(Host.id.in_([r.host_id for r in stale_rows]))
-            )).scalars().all()
-        }
-    return [
-        {
-            "host_id": r.host_id,
-            "host_name": hosts[r.host_id].name if r.host_id in hosts else "?",
-            "last_seen_at": r.last_seen_at,
-        }
-        for r in stale_rows
-    ]
+        async with session.begin():
+            regs = (await session.execute(select(HostAgentRegistration))).scalars().all()
+            watching_since = watch(app, "host_agent_seconds_since_seen").tick(now)
+            if not regs:
+                return
+            host_names = dict((await session.execute(
+                select(Host.id, Host.name).where(Host.id.in_([r.host_id for r in regs]))
+            )).all())
+            # Agents whose last figure was already past the offline window keep
+            # counting through a restart of Steward (#5719).
+            stale_after = agent_stale_after_seconds(app.config)
+            already_silent = set((await session.execute(
+                select(PluginMetricLatest.resource_name).where(
+                    PluginMetricLatest.source_module == "host_agent",
+                    PluginMetricLatest.metric_name == "seconds_since_seen",
+                    PluginMetricLatest.resource_name.in_(list(host_names.values())),
+                    PluginMetricLatest.value > stale_after))).scalars())
+            explained = await announced_stops(session, now, restart_grace_seconds(app.config))
+            batch = MetricBatch()
+            for name, seconds in seconds_since_seen(regs, host_names, now, watching_since,
+                                                    arrivals(app), already_silent, explained):
+                batch.add(source_module="host_agent", resource_name=name,
+                          metric_name="seconds_since_seen", value=seconds, recorded_at=now)
+            await batch.record(session)
 
 
 def make_task(app) -> ScheduledTask:
-    async def _check_stale():
+    async def _record():
         try:
-            stale = await find_stale_registrations(app)
+            await record_seconds_since_seen(app)
         except Exception:
-            logger.exception("host_agent stale check failed")
-            return
-        if stale:
-            logger.info(
-                "host_agent: %d stale agent(s): %s",
-                len(stale),
-                [s["host_name"] for s in stale],
-            )
+            logger.exception("host_agent: recording seconds_since_seen failed")
 
     return ScheduledTask(
-        name="host_agent_stale_check",
-        coro_factory=_check_stale,
+        name="host_agent_seconds_since_seen",
+        coro_factory=_record,
+        interval_seconds=60,
+        run_on_startup=False,
+    )
+
+
+def make_sensor_task(app) -> ScheduledTask:
+    """Re-judges which temperature sensors are thermometers (#5783), hourly
+    and at startup, from the hourly averages: verdicts live in memory, so a
+    restart rebuilds them before the first hour is out."""
+    async def _judge():
+        from . import sensors
+        try:
+            async with app.db_sessionmaker() as session:
+                found = await sensors.refresh(session, datetime.now(timezone.utc))
+            for host, labels in found.items():
+                logger.info("host_agent: %s sensors not counted as temperature: %s",
+                            host, ", ".join(sorted(labels)))
+        except Exception:
+            logger.exception("host_agent: judging temperature sensors failed")
+
+    return ScheduledTask(
+        name="host_agent_sensors",
+        coro_factory=_judge,
+        interval_seconds=3600,
+        run_on_startup=True,
+    )
+
+
+def make_maintenance_task(app) -> ScheduledTask:
+    """Gives up on maintenance jobs past their deadline (#5767), starts due
+    maintenance plans and moves their runs on one host at a time (#5768), and
+    records each host's maintenance_failed for an alert rule. Every move a job
+    itself makes comes from its agent's posts."""
+    async def _tick():
+        from . import cadence, maintenance
+        now = datetime.now(timezone.utc)
+        try:
+            async with app.db_sessionmaker() as session:
+                async with session.begin():
+                    gave_up = await maintenance.tick(session, now)
+                    said = await cadence.tick(session, now)
+                    batch = MetricBatch()
+                    for name, value in (await cadence.failed_hosts(session, now)).items():
+                        batch.add(source_module="host_agent", resource_name=name,
+                                  metric_name="maintenance_failed", value=value, recorded_at=now)
+                    await batch.record(session)
+            if gave_up:
+                logger.warning("host_agent: %d maintenance job(s) timed out", gave_up)
+            for line in said:
+                logger.info("host_agent maintenance: %s", line)
+        except Exception:
+            logger.exception("host_agent: maintenance tick failed")
+
+    return ScheduledTask(
+        name="host_agent_maintenance",
+        coro_factory=_tick,
         interval_seconds=60,
         run_on_startup=False,
     )

plugins/host_agent/sensors.py

@@ -0,0 +1,173 @@
+"""Which of a host's temperature sensors are thermometers (#5783).
+
+A board's spare thermistor inputs often have nothing wired to them, and the
+kernel reports them anyway. No flag says so; their behaviour does. A
+thermometer in a host warms as its CPU warms. An open input reads the chip's
+own reference, which drifts the other way as the chip warms, or holds still:
+VDNT-Util01's SYSTIN ran at r = −0.81 against its CPU package over a week,
+while CPUTIN beside it ran at +0.83 (#5782).
+
+So each sensor is judged against the host's CPU over its hourly averages, by
+a scheduled task, and the verdicts are kept here for ingest and the pages to
+read. Until a host has enough history, or when its CPU never moved enough to
+tell, nothing is judged and routes.counts_as_hottest's value rule decides.
+"""
+from __future__ import annotations
+
+import logging
+import re
+from datetime import datetime, timedelta
+from math import sqrt
+
+logger = logging.getLogger(__name__)
+
+# Hours of history judged, the shared hours a verdict needs, and how far the
+# CPU must have moved in them for "it didn't follow" to mean anything.
+WINDOW = timedelta(days=7)
+MIN_HOURS = 72
+REFERENCE_SWING_C = 10.0
+# Falling as the CPU warms, this strongly, is no thermometer's behaviour.
+# A wired board or drive sensor may barely follow the CPU (r near 0), so
+# only a clear negative counts, not a weak one.
+FALLS_R = -0.5
+# Held within this over days while the CPU moved: a constant, not a reading.
+FLAT_C = 0.5
+
+# The host's CPU, in order of preference: Intel coretemp's package, AMD
+# k10temp's control/die/CCD readings, then the PECI agent.
+_REFERENCES = (re.compile(r"^Package id \d+$", re.I), re.compile(r"^(Tctl|Tdie)$", re.I),
+               re.compile(r"^Tccd\d+$", re.I), re.compile(r"^PECI Agent \d+$", re.I))
+
+# {host name: {sensor label: why it isn't a thermometer}}, from the last refresh.
+_unwired: dict[str, dict[str, str]] = {}
+# {resource: {hour: average}} over the last WINDOW, and the hour the next
+# refresh reads raw samples from. Reading the whole week took over three
+# minutes on the live instance, so it is read once at startup, and each later
+# refresh reads only the hours since the last one.
+_hours: dict[str, dict[datetime, float]] = {}
+_read_from: datetime | None = None
+
+
+def unwired(host_name: str) -> dict[str, str]:
+    """The sensors judged not to be thermometers on a host, with why."""
+    return _unwired.get(host_name, {})
+
+
+def reference_label(labels) -> str | None:
+    """The sensor that reads the host's CPU, or None when it has none."""
+    labels = sorted(labels)
+    for pattern in _REFERENCES:
+        found = [label for label in labels if pattern.match(label)]
+        if found:
+            return found[0]
+    return None
+
+
+def pearson(xs: list[float], ys: list[float]) -> float | None:
+    """Pearson's r, or None when either series doesn't move."""
+    n = len(xs)
+    mx, my = sum(xs) / n, sum(ys) / n
+    sxx = sum((x - mx) ** 2 for x in xs)
+    syy = sum((y - my) ** 2 for y in ys)
+    if sxx == 0 or syy == 0:
+        return None
+    return sum((x - mx) * (y - my) for x, y in zip(xs, ys)) / sqrt(sxx * syy)
+
+
+def judge(reference: dict[datetime, float], sensor: dict[datetime, float]) -> str | None:
+    """Why `sensor` isn't a thermometer, judged against the host's CPU over
+    the hours both have; None when it behaves like one, or when there isn't
+    enough to tell (too few hours, or a CPU that never moved)."""
+    hours = sorted(set(reference) & set(sensor))
+    if len(hours) < MIN_HOURS:
+        return None
+    cpu = [reference[h] for h in hours]
+    got = [sensor[h] for h in hours]
+    if max(cpu) - min(cpu) < REFERENCE_SWING_C:
+        return None
+    if max(got) - min(got) < FLAT_C:
+        return f"holds at {sum(got) / len(got):.0f} °C while the CPU moved {min(cpu):.0f}–{max(cpu):.0f} °C"
+    r = pearson(cpu, got)
+    if r is not None and r <= FALLS_R:
+        return f"falls as the CPU warms (r {r:.2f} over {len(hours)} h)"
+    return None
+
+
+def judge_hosts(rows) -> dict[str, dict[str, str]]:
+    """{host: {label: why}} from hourly rows of (resource, bucket, value_avg),
+    where resource is "<host>:temp:<label>"."""
+    series: dict[str, dict[str, dict[datetime, float]]] = {}
+    for resource, bucket, value in rows:
+        host, sep, label = resource.partition(":temp:")
+        if sep and value is not None:
+            series.setdefault(host, {}).setdefault(label, {})[bucket] = value
+    out: dict[str, dict[str, str]] = {}
+    for host, sensors in series.items():
+        ref = reference_label(sensors)
+        if ref is None:
+            continue
+        for label, points in sensors.items():
+            if label == ref:
+                continue
+            why = judge(sensors[ref], points)
+            if why:
+                out.setdefault(host, {})[label] = why
+    return out
+
+
+def merge_hours(hours: dict[str, dict[datetime, float]], rows, since: datetime) -> None:
+    """Fold (resource, hour, average) rows into `hours`, a newer read of an
+    hour replacing an older one (the hour that was still filling last time),
+    and drop the hours older than `since`."""
+    for resource, bucket, value in rows:
+        if value is not None:
+            hours.setdefault(resource, {})[bucket] = float(value)
+    for resource in list(hours):
+        kept = {h: v for h, v in hours[resource].items() if h >= since}
+        if kept:
+            hours[resource] = kept
+        else:
+            del hours[resource]
+
+
+async def refresh(session, now: datetime) -> dict[str, dict[str, str]]:
+    """Re-judge every host's sensors from the last WINDOW of hourly averages,
+    and keep the verdicts for ingest and the pages.
+
+    Hourly rollups exist only for hours older than the raw window
+    (metrics.retention.raw_days, 7 days by default), so the recent hours are
+    averaged from the raw samples here. Retention deletes the raw rows it
+    rolls up, so the two never cover the same hour. The sensors are named
+    from plugin_metric_latest, so each series is read by its index rather
+    than by a pattern over every host_agent row.
+
+    The first refresh reads the whole WINDOW; later ones read raw samples
+    from the start of the hour the last one ended in, so that hour is
+    re-read whole."""
+    global _unwired, _read_from
+    from sqlalchemy import func, select
+    from steward.models.metrics import PluginMetric as M
+    from steward.models.metrics import PluginMetricHourly as H
+    from steward.models.metrics import PluginMetricLatest as L
+    since = now - WINDOW
+    names = (await session.execute(
+        select(L.resource_name).where(L.source_module == "host_agent", L.metric_name == "temp_c"))).scalars().all()
+    if not names:
+        _unwired = {}
+        return _unwired
+    read_from = since if _read_from is None or _read_from < since else _read_from
+    hour = func.date_trunc("hour", M.recorded_at)
+    rows = list((await session.execute(
+        select(M.resource_name, hour, func.avg(M.value))
+        .where(M.source_module == "host_agent", M.metric_name == "temp_c",
+               M.resource_name.in_(names), M.recorded_at >= read_from)
+        .group_by(M.resource_name, hour))).all())
+    if _read_from is None:
+        rows += (await session.execute(
+            select(H.resource_name, H.bucket, H.value_avg)
+            .where(H.source_module == "host_agent", H.metric_name == "temp_c",
+                   H.resource_name.in_(names), H.bucket >= since))).all()
+    merge_hours(_hours, rows, since)
+    _read_from = now.replace(minute=0, second=0, microsecond=0)
+    _unwired = judge_hosts((r, h, v) for r, pts in _hours.items() for h, v in pts.items())
+    return _unwired

plugins/host_agent/templates/_host_metrics.html

@@ -1,32 +1,15 @@
 {# Current-state fragment for the full-metrics page — polled live via HTMX.
-   Self-contained (defines its own format macros) so each poll re-renders the
-   latest snapshot the server has. #}
-{% macro fmt_bps(v) %}
-{%- if v is none -%}—
-{%- elif v >= 1048576 -%}{{ "%.1f"|format(v / 1048576) }} MB/s
-{%- elif v >= 1024 -%}{{ "%.1f"|format(v / 1024) }} KB/s
-{%- else -%}{{ "%.0f"|format(v) }} B/s
-{%- endif -%}
-{% endmacro %}
-{% macro fmt_bytes(v) %}
-{%- if v is none -%}—
-{%- elif v >= 1125899906842624 -%}{{ "%.1f"|format(v / 1125899906842624) }} PB
-{%- elif v >= 1099511627776 -%}{{ "%.1f"|format(v / 1099511627776) }} TB
-{%- elif v >= 1073741824 -%}{{ "%.1f"|format(v / 1073741824) }} GB
-{%- elif v >= 1048576 -%}{{ "%.0f"|format(v / 1048576) }} MB
-{%- else -%}{{ "%.0f"|format(v / 1024) }} KB
-{%- endif -%}
-{% endmacro %}
-{% macro bar(pct) %}
-{%- set p = pct if pct is not none else 0 -%}
-<div style="height:8px;background:var(--bg-elevated);border-radius:4px;overflow:hidden;border:1px solid var(--border);">
-  <div style="height:100%;width:{{ p|round(1) }}%;background:{% if pct is none %}var(--text-dim){% elif pct < 70 %}var(--green){% elif pct < 90 %}var(--yellow){% else %}var(--red){% endif %};"></div>
-</div>
+   Each poll re-renders the latest snapshot the server has. #}
+{% from "_macros.html" import status_badge %}
+{# A percentage bar graded by Settings → Thresholds for `kind` (cpu, mem, disk). #}
+{% macro bar(pct, kind) %}
+<div class="meter"><div class="meter-fill meter-{{ status_class(pct, kind) }}" style="width:{{ (pct or 0)|round(1) }}%;"></div></div>
 {% endmacro %}
+{% macro gauge(value, kind, text) %}<span class="stat-val tone-{{ status_class(value, kind) }}" style="font-size:1.6rem;">{{ text }}</span>{% endmacro %}
 
 {# ── Identity / metadata ─────────────────────────────────────────────────── #}
 <div class="card" style="display:flex;flex-wrap:wrap;gap:0.5rem 1.25rem;align-items:center;font-size:0.82rem;color:var(--text-muted);">
-  <span>{% if stale %}<span class="badge badge-red">stale</span>{% else %}<span class="badge badge-green">live</span>{% endif %}</span>
+  <span>{{ status_badge('stale' if stale else 'up', 'Stale' if stale else 'Live') }}</span>
   <span><span style="color:var(--text-dim);">Address</span> {{ host.address or "—" }}</span>
   {% if reg %}
   {% if reg.distro %}<span><span style="color:var(--text-dim);">OS</span> {{ reg.distro }}</span>{% endif %}
@@ -34,8 +17,8 @@
   {% if reg.arch %}<span><span style="color:var(--text-dim);">Arch</span> {{ reg.arch }}</span>{% endif %}
   {% if reg.agent_version %}<span><span style="color:var(--text-dim);">Agent</span> v{{ reg.agent_version }}</span>{% endif %}
   {% set up = hostlvl.get('uptime_secs') %}
-  {% if up %}<span><span style="color:var(--text-dim);">Uptime</span> {{ (up // 86400)|int }}d {{ ((up % 86400) // 3600)|int }}h</span>{% endif %}
-  <span><span style="color:var(--text-dim);">Last seen</span> {{ reg.last_seen_at.strftime("%Y-%m-%d %H:%M:%S") ~ " UTC" if reg.last_seen_at else "never" }}</span>
+  {% if up %}<span><span style="color:var(--text-dim);">Uptime</span> {{ up|duration }}</span>{% endif %}
+  <span><span style="color:var(--text-dim);">Last seen</span> {{ reg.last_seen_at|ago if reg.last_seen_at else "never" }}</span>
   {% else %}
   <span>No agent registration found for this host.</span>
   {% endif %}
@@ -50,16 +33,16 @@
   <div class="card" style="margin-bottom:0;">
     <div class="section-title" style="margin-bottom:0.35rem;">CPU</div>
     {% set c = hostlvl.get('cpu_pct') %}
-    <span class="stat-val" style="font-size:1.6rem;color:{% if c is none %}var(--text-dim){% elif c < 70 %}var(--green){% elif c < 90 %}var(--yellow){% else %}var(--red){% endif %};">{% if c is not none %}{{ "%.0f"|format(c) }}%{% else %}—{% endif %}</span>
-    <div style="margin-top:0.5rem;">{{ bar(c) }}</div>
+    {{ gauge(c, 'cpu', ("%.0f%%"|format(c)) if c is not none else "—") }}
+    <div style="margin-top:0.5rem;">{{ bar(c, 'cpu') }}</div>
   </div>
   <div class="card" style="margin-bottom:0;">
     <div class="section-title" style="margin-bottom:0.35rem;">Memory</div>
     {% set mp = hostlvl.get('mem_used_pct') %}
-    <span class="stat-val" style="font-size:1.6rem;color:{% if mp is none %}var(--text-dim){% elif mp < 70 %}var(--green){% elif mp < 90 %}var(--yellow){% else %}var(--red){% endif %};">{% if mp is not none %}{{ "%.0f"|format(mp) }}%{% else %}—{% endif %}</span>
-    <div style="margin-top:0.5rem;">{{ bar(mp) }}</div>
+    {{ gauge(mp, 'mem', ("%.0f%%"|format(mp)) if mp is not none else "—") }}
+    <div style="margin-top:0.5rem;">{{ bar(mp, 'mem') }}</div>
     <div style="margin-top:0.45rem;font-size:0.72rem;color:var(--text-dim);">
-      avail {{ fmt_bytes(hostlvl.get('mem_available_bytes')) }} · cache {{ fmt_bytes(hostlvl.get('mem_cached_bytes')) }} · swap {{ fmt_bytes(hostlvl.get('swap_used_bytes')) }}
+      avail {{ hostlvl.get('mem_available_bytes')|bytes }} · cache {{ hostlvl.get('mem_cached_bytes')|bytes }} · swap {{ hostlvl.get('swap_used_bytes')|bytes }}
     </div>
   </div>
   <div class="card" style="margin-bottom:0;">
@@ -73,22 +56,22 @@
   <div class="card" style="margin-bottom:0;">
     <div class="section-title" style="margin-bottom:0.35rem;">Network</div>
     <div style="font-size:0.95rem;font-variant-numeric:tabular-nums;">
-      <div><span style="color:var(--green);">↓</span> {{ fmt_bps(hostlvl.get('net_rx_bps')) }}</div>
-      <div><span style="color:var(--red);">↑</span> {{ fmt_bps(hostlvl.get('net_tx_bps')) }}</div>
+      <div><span style="color:var(--green);">↓</span> {{ hostlvl.get('net_rx_bps')|rate }}</div>
+      <div><span style="color:var(--red);">↑</span> {{ hostlvl.get('net_tx_bps')|rate }}</div>
     </div>
   </div>
   <div class="card" style="margin-bottom:0;">
     <div class="section-title" style="margin-bottom:0.35rem;">Disk I/O</div>
     <div style="font-size:0.95rem;font-variant-numeric:tabular-nums;">
-      <div><span style="color:var(--text-dim);">rd</span> {{ fmt_bps(hostlvl.get('disk_read_bps')) }}</div>
-      <div><span style="color:var(--text-dim);">wr</span> {{ fmt_bps(hostlvl.get('disk_write_bps')) }}</div>
+      <div><span style="color:var(--text-dim);">rd</span> {{ hostlvl.get('disk_read_bps')|rate }}</div>
+      <div><span style="color:var(--text-dim);">wr</span> {{ hostlvl.get('disk_write_bps')|rate }}</div>
     </div>
   </div>
   {% if hostlvl.get('temp_c_max') is not none %}
   <div class="card" style="margin-bottom:0;">
     <div class="section-title" style="margin-bottom:0.35rem;">Temp (max)</div>
     {% set t = hostlvl.get('temp_c_max') %}
-    <span class="stat-val" style="font-size:1.6rem;color:{% if t < 70 %}var(--green){% elif t < 85 %}var(--yellow){% else %}var(--red){% endif %};">{{ "%.0f"|format(t) }}°C</span>
+    {{ gauge(t, 'temp', "%.0f°C"|format(t)) }}
   </div>
   {% endif %}
   {% if hostlvl.get('psi_mem_some_avg10') is not none %}
@@ -113,7 +96,7 @@
       <div style="display:flex;justify-content:space-between;font-size:0.72rem;color:var(--text-dim);margin-bottom:0.2rem;">
         <span>core {{ idx }}</span><span>{% if pct is not none %}{{ "%.0f"|format(pct) }}%{% else %}—{% endif %}</span>
       </div>
-      {{ bar(pct) }}
+      {{ bar(pct, 'cpu') }}
     </div>
     {% endfor %}
   </div>
@@ -128,9 +111,44 @@
   <div style="margin-bottom:0.6rem;">
     <div style="display:flex;justify-content:space-between;font-size:0.8rem;margin-bottom:0.2rem;">
       <span style="font-family:ui-monospace,monospace;">{{ mount }}</span>
-      <span style="color:var(--text-muted);">{{ fmt_bytes(m.get('disk_used_bytes')) }} / {{ fmt_bytes(m.get('disk_total_bytes')) }}{% if m.get('disk_used_pct') is not none %} · {{ "%.0f"|format(m.get('disk_used_pct')) }}%{% endif %}</span>
+      <span style="color:var(--text-muted);">{{ m.get('disk_used_bytes')|bytes }} / {{ m.get('disk_total_bytes')|bytes }}{% if m.get('disk_used_pct') is not none %} · {{ "%.0f"|format(m.get('disk_used_pct')) }}%{% endif %}</span>
     </div>
-    {{ bar(m.get('disk_used_pct')) }}
+    {{ bar(m.get('disk_used_pct'), 'disk') }}
+  </div>
+  {% endfor %}
+</div>
+{% endif %}
+
+{# ── Drive health (#5789): each drive as the root helper read it with
+   smartctl, hourly; what needs attention first (drives.py judges them). ──── #}
+{% set drive_badge = {"failing": ("crit", "Failing"), "unreadable": ("warn", "Unreadable"),
+                      "worn": ("warn", "Worn"), "watch": ("warn", "Watch"), "ok": ("ok", "OK"),
+                      "unavailable": ("inactive", "No SMART")} %}
+{% if drives or drives_error %}
+<div class="card">
+  <div class="section-title" style="margin-bottom:0.5rem;display:flex;justify-content:space-between;gap:0.75rem;">
+    <span>Drive health</span>
+    {% if drives_at %}<span style="font-weight:400;color:var(--text-dim);font-size:0.72rem;">read {{ drives_at|ago }}</span>{% endif %}
+  </div>
+  {% if drives_error %}
+  <p class="field-hint" style="margin:0;">Not read: {{ drives_error }}.</p>
+  {% endif %}
+  {% for d in drives %}
+  {% set tone, label = drive_badge[d.state] %}
+  <div style="display:flex;justify-content:space-between;align-items:baseline;gap:0.75rem;font-size:0.8rem;padding:0.2rem 0;">
+    <span style="min-width:0;overflow:hidden;text-overflow:ellipsis;white-space:nowrap;" title="{{ d.model }}">
+      {{ status_badge(tone, label) }}
+      <span style="font-family:ui-monospace,monospace;">{{ d.device }}</span>
+      {% if d.model %}<span style="color:var(--text-dim);">{{ d.model }}{% if d.capacity_bytes %} · {{ d.capacity_bytes|bytes }}{% endif %}</span>{% endif %}
+      {% if d.why %}<span style="color:var(--text-muted);"> — {{ d.why }}</span>{% endif %}
+    </span>
+    <span style="color:var(--text-muted);font-variant-numeric:tabular-nums;white-space:nowrap;">
+      {%- set parts = [] -%}
+      {%- if d.get("life_used_pct") is not none %}{% set _ = parts.append("%d%% life used"|format(d.life_used_pct)) %}{% endif -%}
+      {%- if d.get("bad_sectors") is not none %}{% set _ = parts.append("%d bad sectors"|format(d.bad_sectors)) %}{% endif -%}
+      {%- if d.get("temp_c") is not none %}{% set _ = parts.append("%d°C"|format(d.temp_c)) %}{% endif -%}
+      {{ parts|join(" · ") if parts else "—" }}
+    </span>
   </div>
   {% endfor %}
 </div>
@@ -145,7 +163,7 @@
     {% for iface, m in nets.items() %}
     <div style="display:flex;justify-content:space-between;gap:0.75rem;font-size:0.8rem;padding:0.15rem 0;">
       <span style="font-family:ui-monospace,monospace;">{{ iface }}</span>
-      <span style="color:var(--text-muted);font-variant-numeric:tabular-nums;white-space:nowrap;">↓ {{ fmt_bps(m.get('net_rx_bps')) }} · ↑ {{ fmt_bps(m.get('net_tx_bps')) }}</span>
+      <span style="color:var(--text-muted);font-variant-numeric:tabular-nums;white-space:nowrap;">↓ {{ m.get('net_rx_bps')|rate }} · ↑ {{ m.get('net_tx_bps')|rate }}</span>
     </div>
     {% endfor %}
   </div>
@@ -156,7 +174,7 @@
     {% for dev, m in disks_io.items() %}
     <div style="display:flex;justify-content:space-between;gap:0.75rem;font-size:0.8rem;padding:0.15rem 0;">
       <span style="font-family:ui-monospace,monospace;">{{ dev }}</span>
-      <span style="color:var(--text-muted);font-variant-numeric:tabular-nums;white-space:nowrap;">rd {{ fmt_bps(m.get('disk_read_bps')) }} · wr {{ fmt_bps(m.get('disk_write_bps')) }}</span>
+      <span style="color:var(--text-muted);font-variant-numeric:tabular-nums;white-space:nowrap;">rd {{ m.get('disk_read_bps')|rate }} · wr {{ m.get('disk_write_bps')|rate }}</span>
     </div>
     {% endfor %}
   </div>
@@ -172,10 +190,18 @@
     {% for label, c in temps.items() %}
     <div style="display:flex;justify-content:space-between;gap:0.5rem;font-size:0.8rem;padding:0.15rem 0;">
       <span style="white-space:nowrap;overflow:hidden;text-overflow:ellipsis;">{{ label }}</span>
-      <span style="color:{% if c is none %}var(--text-dim){% elif c < 70 %}var(--green){% elif c < 85 %}var(--yellow){% else %}var(--red){% endif %};font-variant-numeric:tabular-nums;">{% if c is not none %}{{ "%.0f"|format(c) }}°C{% else %}—{% endif %}</span>
+      <span class="tone-{{ status_class(c, 'temp') }}" style="font-variant-numeric:tabular-nums;">{% if c is not none %}{{ "%.0f"|format(c) }}°C{% else %}—{% endif %}</span>
     </div>
     {% endfor %}
   </div>
+  {% if unwired_temps %}
+  {# Sensors that aren't thermometers: they run against the CPU or hold a
+     constant (#5783), or read a value no wired sensor gives (#5759). Left
+     out of the host's temperature. Each reason is said once, after its
+     sensors, hottest first within it. #}
+  <p class="field-hint" style="margin:0.5rem 0 0;">Not counted, likely nothing wired to them:
+    {% for why, group in unwired_temps.items()|groupby("1.1") %}{% for label, v in group %}{{ label }} {{ "%.0f"|format(v[0]) }}°C{{ ", " if not loop.last }}{% endfor %} ({{ why }}){{ "; " if not loop.last }}{% endfor %}.</p>
+  {% endif %}
 </div>
 {% endif %}
 

plugins/host_agent/templates/_host_vitals.html

@@ -1,35 +1,42 @@
-{# Host vitals strip — compact CPU/MEM/DISK/LOAD + pressure, live-polled into the
-   host hub. Renders nothing until the agent reports (the Agent panel below then
-   shows provisioning). #}
+{# Host health header — live polled into the host hub: live or stale, the worst
+   reading against Settings → Thresholds, last seen, then CPU / memory / disk /
+   load tiles graded by the same cutoffs (`grades`; CPU, memory and load within
+   the host's own normal for the hour read healthy, and `usual` says so). Renders nothing until the agent
+   reports (the Agent panel below then shows provisioning). #}
+{% from "_macros.html" import status_badge, stat_tile %}
+{% macro pct(v) %}{{ '%.0f%%'|format(v) if v is not none else '—' }}{% endmacro %}
+{% set grades = grades or {} %}
 {% if reporting %}
-{% set lbl = "font-size:0.68rem;color:var(--text-muted);text-transform:uppercase;letter-spacing:0.04em;" %}
-{% macro vital(label, value, text, kind, spark) %}
-<div style="min-width:88px;">
-  <div style="{{ lbl }}">{{ label }}</div>
-  <div style="font-size:1.5rem;font-weight:700;line-height:1.1;{{ threshold_style(value, kind) }}">{{ text }}</div>
-  <div style="margin-top:0.2rem;">{{ spark | safe }}</div>
-</div>
-{% endmacro %}
-<div class="card" style="display:flex;flex-wrap:wrap;align-items:flex-start;gap:1rem 2rem;margin-bottom:1rem;">
-  {{ vital("CPU", cpu, ('%.0f%%'|format(cpu) if cpu is not none else '—'), 'cpu', sparks.cpu) }}
-  {{ vital("Memory", mem, ('%.0f%%'|format(mem) if mem is not none else '—'), 'mem', sparks.mem) }}
-  {{ vital("Disk /", disk_root, ('%.0f%%'|format(disk_root) if disk_root is not none else '—'), 'disk', sparks.disk) }}
-  {{ vital("Load /core", load_per_core, ('%d%%'|format(load_per_core) if load_per_core is not none else '—'), 'load', sparks.load) }}
-
-  <div style="margin-left:auto;display:flex;flex-direction:column;align-items:flex-end;gap:0.3rem;font-size:0.75rem;color:var(--text-muted);text-align:right;">
-    <span>
-      {% if stale %}<span class="badge badge-red">stale</span>{% else %}<span class="badge badge-green">live</span>{% endif %}
-      {% if reg and reg.agent_version %}<span style="margin-left:0.3rem;">v{{ reg.agent_version }}</span>{% endif %}
+<section class="card host-health" aria-label="Host health">
+  <div class="host-health-head">
+    {{ status_badge('stale' if stale else 'up', 'Stale' if stale else 'Live') }}
+    <span class="host-health-summary">
+      {%- if stale %}No report since {{ reg.last_seen_at|ago }}; the readings below are from then.
+      {%- elif worst %}<span class="tone-{{ worst[2] }}">{{ worst[0] }} {{ pct(worst[1]) }}</span> is past its {{ "critical" if worst[2] == "crit" else "warning" }} threshold.
+      {%- elif usual %}Every reading is within its thresholds or its usual level for this hour.
+      {%- else %}Every reading is within its thresholds.{% endif -%}
     </span>
-    {% if psi.cpu is not none or psi.mem is not none or psi.io is not none %}
-    <span title="Pressure stall — % of the last 10s tasks waited for the resource">
-      <span style="color:var(--text-dim);text-transform:uppercase;font-size:0.66rem;letter-spacing:0.04em;">Pressure 10s</span>
-      cpu {{ '%.0f%%'|format(psi.cpu) if psi.cpu is not none else '—' }}
-      · mem {{ '%.0f%%'|format(psi.mem) if psi.mem is not none else '—' }}
-      · io {{ '%.0f%%'|format(psi.io) if psi.io is not none else '—' }}
+    <span class="problem-detail">
+      {%- if not stale %}seen {{ reg.last_seen_at|ago }}{% endif %}
+      {%- if reg.agent_version %}{{ " · " if not stale }}agent v{{ reg.agent_version }}{% endif -%}
+      {%- if clock and clock[0] %} · <span{% if clock[1] in ('warn', 'crit') %} class="tone-{{ clock[1] }}" title="Further from Steward's clock than Settings → Thresholds allows: check what keeps this host's time in sync"{% endif %}>{{ clock[0] }}</span>{% endif -%}
     </span>
-    {% endif %}
-    {% if reg and reg.last_seen_at %}<span>seen {{ reg.last_seen_at.strftime("%H:%M:%S") }} UTC</span>{% endif %}
   </div>
-</div>
+  <div class="stat-strip host-health-tiles" style="margin-bottom:0;">
+    {{ stat_tile("CPU", pct(cpu), state=(grades.cpu or [status_class(cpu, 'cpu')])[0], hint=sparks.cpu|safe) }}
+    {{ stat_tile("Memory", pct(mem), state=(grades.mem or [status_class(mem, 'mem')])[0], hint=sparks.mem|safe) }}
+    {%- set inodes_worse = inodes_root is defined and inodes_root is not none and (disk_root is none or inodes_root > disk_root) %}
+    {{ stat_tile("Disk /", pct(disk_root) ~ (" · inodes " ~ pct(inodes_root) if inodes_worse else ""),
+                 state=(grades.disk or [status_class(disk_root, 'disk')])[0], hint=sparks.disk|safe) }}
+    {{ stat_tile("Load per core", pct(load_per_core), state=(grades.load or [status_class(load_per_core, 'load')])[0], hint=sparks.load|safe) }}
+  </div>
+  {% if usual and not stale %}
+  <p class="field-hint" style="margin-top:0.6rem;">{{ usual | join("; ") | capitalize }}.</p>
+  {% endif %}
+  {% if psi.cpu is not none or psi.mem is not none or psi.io is not none %}
+  <p class="field-hint" style="margin-top:0.6rem;" title="Pressure stall: the share of the last 10 s that tasks waited for the resource">
+    Pressure (10 s): cpu {{ pct(psi.cpu) }} · mem {{ pct(psi.mem) }} · io {{ pct(psi.io) }}
+  </p>
+  {% endif %}
+</section>
 {% endif %}

plugins/host_agent/templates/host_detail.html

@@ -54,30 +54,30 @@
     interaction: { mode: "index", intersect: false },
     elements: { point: { radius: 0 } },
     scales: {
-      x: { type: "linear", ticks: { callback: (v) => fmtTime(v), maxTicksLimit: 8, color: "#8a8a92", font: { size: 10 } }, grid: { color: "#30303a" } },
-      y: { beginAtZero: true, max: ymax, ticks: { color: "#8a8a92", font: { size: 10 } }, grid: { color: "#30303a" } },
+      x: { type: "linear", ticks: { callback: (v) => fmtTime(v), maxTicksLimit: 8, font: { size: 10 } } },
+      y: { beginAtZero: true, max: ymax, ticks: { font: { size: 10 } } },
     },
     plugins: {
-      legend: { labels: { color: "#b8b8b0", boxWidth: 10, font: { size: 11 } } },
+      legend: { labels: { boxWidth: 10, font: { size: 11 } } },
       tooltip: { callbacks: { title: (items) => items.length ? fmtTime(items[0].parsed.x) : "" } },
     },
   });
   const defs = {
     "chart-util": { ymax: 100, ds: [
-      { key: "cpu_pct", label: "CPU %", color: "#c8a840" },
-      { key: "mem_used_pct", label: "Mem %", color: "#4aa86a" },
-      { key: "disk_used_pct_worst", label: "Disk % (worst)", color: "#c87840" },
+      { key: "cpu_pct", label: "CPU %", color: stewardColor("--gold") },
+      { key: "mem_used_pct", label: "Mem %", color: stewardColor("--green") },
+      { key: "disk_used_pct_worst", label: "Disk % (worst)", color: stewardColor("--yellow") },
     ] },
     "chart-net": { ymax: undefined, ds: [
-      { key: "net_rx_bps", label: "Net RX", color: "#4aa86a" },
-      { key: "net_tx_bps", label: "Net TX", color: "#c84048" },
-      { key: "disk_read_bps", label: "Disk read", color: "#6aa0d0" },
-      { key: "disk_write_bps", label: "Disk write", color: "#c8a840" },
+      { key: "net_rx_bps", label: "Net RX", color: stewardColor("--green") },
+      { key: "net_tx_bps", label: "Net TX", color: stewardColor("--red") },
+      { key: "disk_read_bps", label: "Disk read", color: stewardColor("--blue") },
+      { key: "disk_write_bps", label: "Disk write", color: stewardColor("--gold") },
     ] },
     "chart-pressure": { ymax: undefined, ds: [
-      { key: "load_1m", label: "Load 1m", color: "#c8a840" },
-      { key: "psi_mem_some_avg10", label: "Mem PSI %", color: "#c84048" },
-      { key: "temp_c_max", label: "Temp °C", color: "#c87840" },
+      { key: "load_1m", label: "Load 1m", color: stewardColor("--gold") },
+      { key: "psi_mem_some_avg10", label: "Mem PSI %", color: stewardColor("--red") },
+      { key: "temp_c_max", label: "Temp °C", color: stewardColor("--yellow") },
     ] },
   };
   const charts = {};

plugins/host_agent/templates/install.sh.j2

@@ -11,7 +11,9 @@ AGENT_VERSION="{{ agent_version }}"
 AGENT_USER="steward-agent"
 AGENT_DIR="/usr/local/lib/steward-agent"
 CONF_FILE="/etc/steward-agent.conf"
-UNIT_FILE="/etc/systemd/system/steward-agent.service"
+UNIT_DIR="/etc/systemd/system"
+UNITS="steward-agent.service steward-maint.service steward-maint.path"
+FILES_URL="${STEWARD_URL}/plugins/host_agent/files"
 
 # ── preflight ────────────────────────────────────────────────────────────────
 [ "$(id -u)" = "0" ] || { echo "must run as root (use sudo)"; exit 1; }
@@ -20,9 +22,10 @@ command -v python3   >/dev/null 2>&1 || { echo "python3 not found — install py
 
 # Handle --uninstall
 if [ "${1:-}" = "--uninstall" ]; then
-    systemctl disable --now steward-agent.service 2>/dev/null || true
-    rm -f "$UNIT_FILE" "$CONF_FILE"
-    rm -rf "$AGENT_DIR"
+    systemctl disable --now steward-maint.path steward-agent.service 2>/dev/null || true
+    for u in $UNITS; do rm -f "$UNIT_DIR/$u"; done
+    rm -f "$CONF_FILE"
+    rm -rf "$AGENT_DIR" /var/lib/steward-agent /var/lib/steward-maint
     systemctl daemon-reload
     # keep the system user — removing it risks orphaned files elsewhere
     echo "uninstalled"
@@ -36,7 +39,7 @@ fi
 
 # ── drop the agent file ──────────────────────────────────────────────────────
 mkdir -p "$AGENT_DIR"
-curl -sSL "${STEWARD_URL}/plugins/host_agent/agent.py" -o "$AGENT_DIR/agent.py"
+curl -fsSL "${STEWARD_URL}/plugins/host_agent/agent.py" -o "$AGENT_DIR/agent.py"
 chmod 0755 "$AGENT_DIR/agent.py"
 chown root:root "$AGENT_DIR/agent.py"
 
@@ -45,35 +48,38 @@ cat > "$CONF_FILE" <<EOF
 url = $STEWARD_URL
 token = $AGENT_TOKEN
 interval_seconds = 30
+# Container logs are collected by default. To opt this host out entirely:
+#   docker_logs_enabled = false
+# To skip specific noisy containers (comma-separated names):
+#   docker_log_exclude = watchtower, some-chatty-service
 EOF
 chown "root:$AGENT_USER" "$CONF_FILE"
 chmod 0640 "$CONF_FILE"
 
-# ── write systemd unit ───────────────────────────────────────────────────────
-cat > "$UNIT_FILE" <<EOF
-[Unit]
-Description=Steward host agent
-After=network-online.target
-Wants=network-online.target
-
-[Service]
-Type=simple
-User=$AGENT_USER
-ExecStart=/usr/bin/python3 $AGENT_DIR/agent.py
-Restart=on-failure
-RestartSec=10
-NoNewPrivileges=yes
-ProtectSystem=strict
-ProtectHome=yes
-PrivateTmp=yes
-ReadOnlyPaths=/proc /sys
-
-[Install]
-WantedBy=multi-user.target
-EOF
+# ── the maintenance helper and the units (#5766) ─────────────────────────────
+# One copy, served by Steward, the same the playbooks install. The agent may
+# write only its request directory; the helper (root) and its results don't
+# belong to it.
+curl -fsSL "${FILES_URL}/maint.py" -o "$AGENT_DIR/maint.py"
+chmod 0755 "$AGENT_DIR/maint.py"
+chown root:root "$AGENT_DIR/maint.py"
+# Its packages (smartctl for drive health, #5789), with whichever package
+# manager this host has; one that can't be installed is said, never fatal.
+curl -fsSL "${FILES_URL}/install-deps.sh" -o "$AGENT_DIR/install-deps.sh"
+chmod 0755 "$AGENT_DIR/install-deps.sh"
+chown root:root "$AGENT_DIR/install-deps.sh"
+"$AGENT_DIR/install-deps.sh" || true
+install -d -o root -g root -m 0755 /var/lib/steward-agent /var/lib/steward-maint /var/lib/steward-maint/results
+install -d -o "$AGENT_USER" -g "$AGENT_USER" -m 0700 /var/lib/steward-agent/requests
+for u in $UNITS; do
+    curl -fsSL "${FILES_URL}/$u" -o "$UNIT_DIR/$u"
+    chmod 0644 "$UNIT_DIR/$u"
+done
 
 systemctl daemon-reload
-systemctl enable --now steward-agent.service
+systemctl enable --now steward-maint.path
+systemctl enable steward-agent.service
+systemctl restart steward-agent.service
 
 echo
 echo "Steward host agent $AGENT_VERSION installed and running."

plugins/host_agent/templates/maintenance.html

@@ -0,0 +1,143 @@
+{% extends "base.html" %}
+{% from "_macros.html" import crumbs, empty_state, page_header %}
+{% block title %}Maintenance — Steward{% endblock %}
+{% block breadcrumb %}{{ crumbs([("Hosts", "/hosts/"), ("Maintenance", "")]) }}{% endblock %}
+{% block content %}
+{# A group's updates and reboots on a weekly cadence, one host at a time
+   (#5768). What needs attention first: a run going now, a run that failed;
+   then the plans with their recent runs. #}
+{% macro run_hosts(rows) -%}
+<div style="display:flex;flex-direction:column;gap:0.15rem;font-size:0.82rem;">
+  {% for r in rows %}
+  <div style="display:flex;gap:0.6rem;align-items:baseline;" {% if r.detail %}title="{{ r.detail }}"{% endif %}>
+    <a href="/hosts/{{ r.host_id }}" style="min-width:10rem;">{{ r.name }}</a>
+    <span class="{{ 'tone-warn' if r.bad else '' }}" style="{{ '' if r.bad or r.busy else 'color:var(--text-muted);' }}">{{ r.text }}</span>
+  </div>
+  {% endfor %}
+</div>
+{%- endmacro %}
+
+{{ page_header("Maintenance") }}
+<p style="color:var(--text-muted);font-size:0.84rem;margin:-0.4rem 0 1.25rem;">
+  Each plan updates an inventory group's hosts once a week, one host at a time. A host reboots
+  only when its update needs it, unless the plan says otherwise; a swarm node is drained first
+  and made active again after. The next host starts once the last is back in service. A failure
+  stops the run, and the host's <em>Maintenance failed</em> figure can drive an alert rule.
+</p>
+
+{% if live %}
+{% for l in live %}
+<div class="card">
+  <h3 style="margin-bottom:0.5rem;">Running: {{ l.plan }} <span style="font-size:0.78rem;color:var(--text-muted);font-weight:normal;">· started {{ l.run.started_at|ago }}{% if l.run.manual %} · run now{% endif %}</span></h3>
+  {{ run_hosts(l.rows) }}
+</div>
+{% endfor %}
+{% endif %}
+
+{% for f in failed %}
+<div class="card" style="border-color:color-mix(in srgb,var(--yellow) 40%,var(--border));">
+  <span class="tone-warn">Failed: {{ f.plan }}</span>
+  <span style="font-size:0.82rem;color:var(--text-muted);">· {{ f.run.finished_at|ago }} · {{ f.run.detail or "no detail" }}</span>
+</div>
+{% endfor %}
+
+{% if plans %}
+{% for p in plans %}
+<div class="card">
+  <div style="display:flex;align-items:baseline;justify-content:space-between;gap:1rem;flex-wrap:wrap;">
+    <h3 style="margin-bottom:0.2rem;">{{ p.plan.name }}{% if not p.plan.enabled %} <span style="font-size:0.78rem;color:var(--text-muted);font-weight:normal;">· paused</span>{% endif %}</h3>
+    {% if session.user_role == 'admin' %}
+    <div style="display:flex;gap:0.4rem;">
+      <form method="post" action="/plugins/host_agent/maintenance/plans/{{ p.plan.id }}/run" style="margin:0;"
+            data-confirm="Run {{ p.plan.name }} now? Its {{ p.hosts }} host{{ 's' if p.hosts != 1 }} are updated one at a time; the schedule keeps its time.">{{ csrf_field() }}
+        <button type="submit" class="btn btn-sm">Run now</button>
+      </form>
+      <form method="post" action="/plugins/host_agent/maintenance/plans/{{ p.plan.id }}/toggle" style="margin:0;">{{ csrf_field() }}
+        <button type="submit" class="btn btn-sm btn-ghost">{{ 'Pause' if p.plan.enabled else 'Resume' }}</button>
+      </form>
+      <form method="post" action="/plugins/host_agent/maintenance/plans/{{ p.plan.id }}/delete" style="margin:0;"
+            data-confirm="Delete {{ p.plan.name }} and its run history? Hosts' own job history stays.">{{ csrf_field() }}
+        <button type="submit" class="btn btn-sm btn-danger">Delete</button>
+      </form>
+    </div>
+    {% endif %}
+  </div>
+  <div style="font-size:0.82rem;color:var(--text-muted);margin-bottom:0.6rem;">
+    Group <a href="/hosts/?group={{ p.group|urlencode }}">{{ p.group }}</a> ({{ p.hosts }} host{{ 's' if p.hosts != 1 }})
+    · {{ p.when }}
+    · reboot {{ {'if_required': 'only when required', 'always': 'always', 'never': 'never'}[p.plan.reboot_policy] }}
+    {% if p.next %}· next {{ p.next|when }}{% endif %}
+  </div>
+  {% for r in p.runs %}
+  <details {% if loop.first %}open{% endif %} style="margin-top:0.3rem;">
+    <summary style="cursor:pointer;font-size:0.82rem;">
+      <span class="{{ 'tone-warn' if r.run.state == 'failed' else '' }}">{{ {'running': 'Running', 'done': 'Done', 'failed': 'Failed', 'stopped': 'Stopped'}[r.run.state] }}</span>
+      <span style="color:var(--text-muted);">· {{ r.run.started_at|when }}{% if r.run.manual %} · run now{% endif %}{% if r.run.detail %} · {{ r.run.detail|truncate(100) }}{% endif %}</span>
+    </summary>
+    <div style="margin:0.3rem 0 0.4rem 1rem;">{{ run_hosts(r.rows) }}</div>
+  </details>
+  {% else %}
+  <p style="font-size:0.82rem;color:var(--text-dim);margin:0;">No runs yet.</p>
+  {% endfor %}
+</div>
+{% endfor %}
+{% else %}
+<div class="card">{{ empty_state("No maintenance plans yet." ~ (" Add one below." if session.user_role == 'admin' else "")) }}</div>
+{% endif %}
+
+{% if session.user_role == 'admin' %}
+<div class="card">
+  <h3 style="margin-bottom:0.6rem;">Add a plan</h3>
+  {% if groups %}
+  <form method="post" action="/plugins/host_agent/maintenance/plans"
+        style="display:flex;gap:0.75rem;align-items:flex-end;flex-wrap:wrap;">{{ csrf_field() }}
+    <div class="form-group" style="margin-bottom:0;">
+      <label>Name</label>
+      <input type="text" name="name" required maxlength="128" placeholder="Docker hosts, weekly" style="width:13rem;">
+    </div>
+    <div class="form-group" style="margin-bottom:0;">
+      <label>Group</label>
+      <select name="group_id" required>
+        {% for g in groups %}<option value="{{ g.id }}">{{ g.name }}</option>{% endfor %}
+      </select>
+    </div>
+    <div class="form-group" style="margin-bottom:0;">
+      <label>Day</label>
+      <select name="weekday">
+        {% for d in weekdays %}<option value="{{ loop.index0 }}" {% if loop.index0 == 6 %}selected{% endif %}>{{ d }}</option>{% endfor %}
+      </select>
+    </div>
+    <div class="form-group" style="margin-bottom:0;">
+      <label>Time</label>
+      <input type="time" name="at_time" value="03:00" required style="width:7rem;">
+    </div>
+    <div class="form-group" style="margin-bottom:0;">
+      <label>Time zone</label>
+      <input type="text" name="time_zone" value="{{ default_zone }}" required style="width:11rem;">
+    </div>
+    <div class="form-group" style="margin-bottom:0;">
+      <label>Window (min)</label>
+      <input type="number" name="window_minutes" value="240" min="30" max="1440" style="width:6rem;">
+    </div>
+    <div class="form-group" style="margin-bottom:0;">
+      <label>Reboot</label>
+      <select name="reboot_policy">
+        <option value="if_required" selected>Only when required</option>
+        <option value="always">Always</option>
+        <option value="never">Never</option>
+      </select>
+    </div>
+    <button type="submit" class="btn">Add plan</button>
+  </form>
+  <p style="font-size:0.78rem;color:var(--text-dim);margin:0.6rem 0 0;">
+    Hosts whose agent has no maintenance helper are skipped and listed; run Update agent on them.
+    No host is started once the window has passed.
+  </p>
+  {% else %}
+  <p style="font-size:0.85rem;color:var(--text-dim);margin:0;">
+    Plans work on inventory groups. Add one under <a href="/ansible/inventory/">Ansible → Inventory</a>.
+  </p>
+  {% endif %}
+</div>
+{% endif %}
+{% endblock %}

plugins/host_agent/templates/panel.html

@@ -4,9 +4,9 @@
   <div style="display:flex;align-items:baseline;justify-content:space-between;margin-bottom:0.75rem;gap:1rem;flex-wrap:wrap;">
     <h3 class="section-title" style="margin-bottom:0;">Agent</h3>
     {% if reporting %}
-    <span style="font-size:0.78rem;color:{{ 'var(--yellow)' if stale else 'var(--green)' }};">
+    <span style="font-size:0.78rem;" class="tone-{{ 'warn' if stale else 'ok' }}">
       {{ 'stale' if stale else 'reporting' }}{% if reg.agent_version %} · v{{ reg.agent_version }}{% endif %}
-      · last seen {{ reg.last_seen_at.strftime('%Y-%m-%d %H:%M') }}
+      · last seen {{ reg.last_seen_at|ago }}
     </span>
     {% elif reg %}
     <span style="font-size:0.78rem;color:var(--text-muted);">pending — no check-in yet</span>
@@ -15,52 +15,119 @@
 
   {% if reporting %}
   {# Reporting: live vitals are shown by the vitals strip above; this panel is
-     lifecycle/management only. #}
-  <div style="display:flex;gap:0.5rem;flex-wrap:wrap;align-items:center;">
-    <a href="/plugins/host_agent/{{ host.id }}/" class="btn btn-sm btn-ghost">Full metrics →</a>
-    {% if session.user_role == 'admin' %}
-    {% if ansible_available and target %}
-    <form method="post" action="/plugins/host_agent/update" style="margin:0;">
-      <input type="hidden" name="inventory_scope" value="{{ scope }}">
-      <button type="submit" class="btn btn-sm" title="Refresh agent.py + restart (token preserved)">Update agent</button>
-    </form>
+     the way to the full metrics, with the lifecycle actions folded under
+     "Manage agent" (#5278). #}
+  <a href="/plugins/host_agent/{{ host.id }}/" class="btn btn-sm btn-ghost">Full metrics →</a>
+
+  {# What the host has waiting, from the agent's hourly check (#5765): a
+     needed reboot first, then the updates. #}
+  {% if reg.updates_checked_at or reg.reboot_required or maint_job %}
+  <div style="margin-top:0.8rem;font-size:0.82rem;display:flex;flex-direction:column;gap:0.2rem;">
+    {# Steward's own maintenance job on this host (#5767): a running one first. #}
+    {% if maint_job %}
+    {% set running = maint_job.state not in ('done', 'failed', 'timed_out') %}
+    <div class="{{ 'tone-warn' if running or maint_job.state != 'done' else '' }}" title="{{ maint_job.detail or '' }}">
+      <a href="/plugins/host_agent/maintenance" style="color:inherit;">Maintenance</a>: {{ maint_summary }}
+      <span style="color:var(--text-muted);">· {% if running %}since {{ maint_job.state_at|ago }}{% else %}{{ maint_job.finished_at|ago }}{% endif %}{% if maint_job.state in ('failed', 'timed_out') and maint_job.detail %} · {{ maint_job.detail|truncate(90) }}{% endif %}</span>
+    </div>
     {% endif %}
-    <form method="post" action="/plugins/host_agent/fleet/{{ host.id }}/rotate-token" style="margin:0;"
-          onsubmit="return confirm('Rotate token? The agent stops reporting until reinstalled/updated with the new token.');">
-      <button type="submit" class="btn btn-sm btn-ghost">Rotate token</button>
-    </form>
-    <form method="post" action="/plugins/host_agent/fleet/{{ host.id }}/delete" style="margin:0;"
-          onsubmit="return confirm('Remove this agent registration? Metrics stop until re-registered.');">
-      <button type="submit" class="btn btn-sm btn-danger">Remove</button>
-    </form>
+    {% if reg.reboot_required %}
+    <div class="tone-warn" title="{{ reg.reboot_packages or '' }}">
+      Reboot required{% if reg.reboot_required_since %} since {{ reg.reboot_required_since|ago }}{% endif %}{% if reg.reboot_packages %}
+      <span style="color:var(--text-muted);">· {{ reg.reboot_packages|truncate(80) }}</span>{% endif %}
+    </div>
+    {% endif %}
+    {% if reg.updates_checked_at %}
+    <div>
+      {% if reg.updates_error %}<span class="tone-warn">Update check failed: {{ reg.updates_error }}</span>
+      {% elif reg.updates_pending %}<span class="{{ 'tone-warn' if reg.updates_security else '' }}">{{ reg.updates_pending }} update{{ 's' if reg.updates_pending != 1 }} waiting{% if reg.updates_security %} ({{ reg.updates_security }} security){% endif %}</span>
+      {% else %}Up to date{% endif %}
+      <span style="color:var(--text-muted);">· checked {{ reg.updates_checked_at|ago }}</span>
+    </div>
     {% endif %}
   </div>
-  {% if ansible_available and not target %}
-  <p style="color:var(--text-dim);font-size:0.8rem;margin:0.6rem 0 0;">
-    Link an Ansible target (above) to enable one-click agent updates.
-  </p>
   {% endif %}
 
-  {% if session.user_role == 'admin' and ansible_available and target and managed_key_set %}
+  {# What the agent said happened while it was away, last 7 days (#5756). #}
+  {% if downtime %}
+  <div style="margin-top:0.8rem;">
+    <div style="font-size:0.78rem;color:var(--text-muted);margin-bottom:0.3rem;">Downtime, last 7 days</div>
+    {% for d in downtime %}
+    <div style="display:flex;gap:0.6rem;align-items:baseline;font-size:0.82rem;padding:0.15rem 0;">
+      <span class="{{ 'tone-warn' if d.ongoing or d.kind == 'unclean' else '' }}" style="min-width:12rem;">{{ d.label }}</span>
+      <span style="color:var(--text-muted);">{{ d.started_at|when }}</span>
+      <span style="color:var(--text-muted);margin-left:auto;">{{ 'going on, ' if d.ongoing }}{{ d.seconds|duration }}</span>
+    </div>
+    {% endfor %}
+  </div>
+  {% endif %}
+
+  {% if session.user_role == 'admin' %}
   <details style="margin-top:0.7rem;">
-    <summary style="cursor:pointer;font-size:0.8rem;color:var(--text-muted);">Re-provision (reinstall the steward account + managed key, then the agent)</summary>
-    <p style="font-size:0.78rem;color:var(--text-dim);margin:0.5rem 0;">
-      Use after regenerating the managed key, or if SSH auth as <code>steward</code> breaks.
-      Connects over a one-time bootstrap user + password (not stored).
+    <summary style="cursor:pointer;font-size:0.82rem;color:var(--text-muted);">Manage agent</summary>
+    <div style="display:flex;gap:0.5rem;flex-wrap:wrap;align-items:center;margin-top:0.6rem;">
+      {% if ansible_available and target %}
+      <form method="post" action="/plugins/host_agent/update" style="margin:0;">{{ csrf_field() }}
+        <input type="hidden" name="inventory_scope" value="{{ scope }}">
+        <button type="submit" class="btn btn-sm" title="Refresh agent.py and restart it; the token is kept">Update agent</button>
+      </form>
+      {% endif %}
+      <form method="post" action="/plugins/host_agent/fleet/{{ host.id }}/rotate-token" style="margin:0;"
+            data-confirm="Rotate token? The agent stops reporting until reinstalled/updated with the new token.">{{ csrf_field() }}
+        <button type="submit" class="btn btn-sm btn-ghost">Rotate token</button>
+      </form>
+      <form method="post" action="/plugins/host_agent/fleet/{{ host.id }}/delete" style="margin:0;"
+            data-confirm="Unregister this host's agent? Its token stops working and metrics stop until it is registered again. The agent itself stays installed on the host.">{{ csrf_field() }}
+        <button type="submit" class="btn btn-sm btn-danger">Unregister agent</button>
+      </form>
+    </div>
+    {% if ansible_available and not target %}
+    <p style="color:var(--text-dim);font-size:0.8rem;margin:0.6rem 0 0;">
+      Link an Ansible target (below) to update the agent from here.
     </p>
-    <form method="post" action="/plugins/host_agent/provision"
-          style="display:flex;gap:0.6rem;align-items:flex-end;flex-wrap:wrap;">
-      <input type="hidden" name="inventory_scope" value="{{ scope }}">
-      <div class="form-group" style="margin:0;">
-        <label>Bootstrap user</label>
-        <input type="text" name="bootstrap_user" required placeholder="root" style="width:8rem;" autocomplete="off">
-      </div>
-      <div class="form-group" style="margin:0;">
-        <label>Bootstrap password</label>
-        <input type="password" name="bootstrap_password" required style="width:10rem;" autocomplete="new-password">
-      </div>
-      <button type="submit" class="btn btn-sm">Re-provision</button>
+    {% endif %}
+    {# Update now (#5767): the agent's helper updates the host and reboots it
+       only if the update needs it, draining its swarm node first. #}
+    {% if maint_blocker %}
+    <p style="font-size:0.8rem;margin:0.6rem 0 0;color:var(--text-dim);">Can't update from here: {{ maint_blocker }}.</p>
+    {% else %}
+    <form method="post" action="/plugins/host_agent/fleet/{{ host.id }}/maintain"
+          style="display:flex;gap:0.6rem;align-items:center;flex-wrap:wrap;margin:0.6rem 0 0;"
+          data-confirm="Update {{ host.name }} now? It reboots only if the update needs it, after its swarm node is drained.">{{ csrf_field() }}
+      <button type="submit" class="btn btn-sm">Update now</button>
+      <label style="font-size:0.8rem;color:var(--text-muted);display:flex;gap:0.3rem;align-items:center;">
+        <input type="checkbox" name="reboot_policy" value="never"> never reboot
+      </label>
     </form>
+    {% endif %}
+    {# The root helper that carries out scheduled updates and reboots (#5766). #}
+    <p style="font-size:0.8rem;margin:0.6rem 0 0;color:var(--text-dim);">
+      {% if reg.maint_helper %}Maintenance helper installed: Steward can update and reboot this host.
+      {% else %}No maintenance helper yet: Update agent installs it, so Steward can update and reboot this host.{% endif %}
+    </p>
+
+    {% if ansible_available and target and managed_key_set %}
+    <details style="margin-top:0.7rem;">
+      <summary style="cursor:pointer;font-size:0.8rem;color:var(--text-muted);">Re-provision (reinstall the steward account + managed key, then the agent)</summary>
+      <p style="font-size:0.78rem;color:var(--text-dim);margin:0.5rem 0;">
+        Use after regenerating the managed key, or if SSH auth as <code>steward</code> breaks.
+        Connects over a one-time bootstrap user + password (not stored).
+      </p>
+      <form method="post" action="/plugins/host_agent/provision"
+            style="display:flex;gap:0.6rem;align-items:flex-end;flex-wrap:wrap;">{{ csrf_field() }}
+        <input type="hidden" name="inventory_scope" value="{{ scope }}">
+        <div class="form-group" style="margin:0;">
+          <label>Bootstrap user</label>
+          <input type="text" name="bootstrap_user" required placeholder="root" style="width:8rem;" autocomplete="off">
+        </div>
+        <div class="form-group" style="margin:0;">
+          <label>Bootstrap password</label>
+          <input type="password" name="bootstrap_password" required style="width:10rem;" autocomplete="new-password">
+        </div>
+        <button type="submit" class="btn btn-sm">Re-provision</button>
+      </form>
+    </details>
+    {% endif %}
   </details>
   {% endif %}
 
@@ -86,8 +153,7 @@
 
   {% elif not ansible_available %}
   <p style="color:var(--text-dim);font-size:0.85rem;">
-    Ansible isn't available, so the agent can't be deployed from here. Install it manually with the
-    <a href="/plugins/host_agent/fleet/">curl install command</a>.
+    Ansible isn't available, so the agent can't be deployed from here. Install it manually instead (below).
   </p>
 
   {% elif not managed_key_set %}
@@ -96,7 +162,7 @@
               border-radius:6px;padding:0.7rem 0.9rem;display:flex;align-items:center;gap:0.8rem;flex-wrap:wrap;">
     <span style="color:var(--yellow);">⚠</span>
     <span style="font-size:0.84rem;flex:1;min-width:13rem;">No managed SSH key yet — Steward needs one to log into hosts.</span>
-    <form method="post" action="/settings/ansible/generate-key" style="margin:0;">
+    <form method="post" action="/settings/ansible/generate-key" style="margin:0;">{{ csrf_field() }}
       <input type="hidden" name="next" value="/hosts/{{ host.id }}">
       <button type="submit" class="btn btn-sm">Generate managed key</button>
     </form>
@@ -115,7 +181,7 @@
   </p>
   {# Provision: brand-new host (creates steward account + key over a one-time password) #}
   <form method="post" action="/plugins/host_agent/provision"
-        style="display:flex;gap:0.6rem;align-items:flex-end;flex-wrap:wrap;margin-bottom:0.75rem;">
+        style="display:flex;gap:0.6rem;align-items:flex-end;flex-wrap:wrap;margin-bottom:0.75rem;">{{ csrf_field() }}
     <input type="hidden" name="inventory_scope" value="{{ scope }}">
     <div style="font-size:0.78rem;color:var(--text-muted);width:100%;">Provision (first contact — fresh host):</div>
     <div class="form-group" style="margin:0;">
@@ -129,16 +195,26 @@
     <button type="submit" class="btn btn-sm">Provision</button>
   </form>
   {# Install: host already has the steward account (managed key works) #}
-  <form method="post" action="/plugins/host_agent/deploy" style="display:flex;gap:0.6rem;align-items:center;">
+  <form method="post" action="/plugins/host_agent/deploy" style="display:flex;gap:0.6rem;align-items:center;">{{ csrf_field() }}
     <input type="hidden" name="inventory_scope" value="{{ scope }}">
     <span style="font-size:0.78rem;color:var(--text-muted);">Already provisioned?</span>
     <button type="submit" class="btn btn-sm btn-ghost">Install agent (managed key)</button>
   </form>
   {% endif %}
 
+  {% if not reg and session.user_role == 'admin' %}
+  {# A manual install: no Ansible, no SSH. Issues this host's token and shows
+     the one-line install command on the fleet page, once (#5282). #}
+  <form method="post" action="/plugins/host_agent/fleet/{{ host.id }}/register"
+        style="display:flex;gap:0.6rem;align-items:center;flex-wrap:wrap;margin:0.75rem 0 0;">{{ csrf_field() }}
+    <span style="font-size:0.78rem;color:var(--text-muted);">Or by hand on the host:</span>
+    <button type="submit" class="btn btn-sm btn-ghost">Install manually</button>
+  </form>
+  {% endif %}
+
   {% if reg and session.user_role == 'admin' %}
   <form method="post" action="/plugins/host_agent/fleet/{{ host.id }}/delete" style="margin:0.75rem 0 0;"
-        onsubmit="return confirm('Clear this pending registration? Its token is discarded.');">
+        data-confirm="Clear this pending registration? Its token is discarded.">{{ csrf_field() }}
     <button type="submit" class="btn btn-sm btn-ghost">Clear pending registration</button>
   </form>
   {% endif %}

plugins/host_agent/templates/settings_list.html

@@ -1,5 +1,5 @@
 {% extends "base.html" %}
-{% from "_macros.html" import crumbs %}
+{% from "_macros.html" import crumbs, status_badge %}
 {% block title %}Agent fleet — Steward{% endblock %}
 {% block breadcrumb %}{{ crumbs([("Hosts", "/hosts/"), ("Agent fleet", "")]) }}{% endblock %}
 {% block content %}
@@ -8,11 +8,12 @@
 </div>
 <p style="color:var(--text-muted);font-size:0.84rem;margin:-0.4rem 0 1.25rem;">
   Bulk agent operations across your inventory. For a single host, use its
-  <a href="/hosts/">host page</a> — provisioning and metrics live there.
+  <a href="/hosts/">host page</a>: provisioning, a manual install command and
+  metrics live there. New hosts are added with <a href="/hosts/?add=1">Add host</a>.
 </p>
 
 {% if new_token %}
-<div class="card" style="background:var(--accent-bg);">
+<div class="card" style="background:var(--fs-accent-faint);">
   <h3>New token — copy this install command now</h3>
   <p style="font-size:0.82rem;color:var(--text-muted);">
     This is the only time the raw token is shown. Rotate if you lose it.
@@ -21,19 +22,43 @@
 </div>
 {% endif %}
 
-<div class="card">
-  <form method="post" action="/plugins/host_agent/fleet/add-host"
-        style="display:flex;gap:0.75rem;align-items:flex-end;flex-wrap:wrap;">
-    <div class="form-group" style="margin-bottom:0;">
-      <label>Host name</label>
-      <input type="text" name="name" required>
-    </div>
-    <div class="form-group" style="margin-bottom:0;">
-      <label>Address (optional)</label>
-      <input type="text" name="address">
-    </div>
-    <button type="submit" class="btn">Add host</button>
-  </form>
+<div class="card-flush" style="margin-bottom:1.5rem;">
+  <div class="table-wrap">
+  <table class="table">
+    <thead>
+      <tr>
+        <th>Status</th><th>Host</th><th>Last seen</th><th>Agent version</th><th>Reported IP</th><th>Distro</th>
+        <th class="td-actions">Actions</th>
+      </tr>
+    </thead>
+    <tbody>
+      {% for item in registrations %}
+      <tr>
+        <td>{{ status_badge(item.state, {'up': 'Reporting', 'stale': 'Silent', 'pending': 'Never reported'}[item.state]) }}</td>
+        <td>{% if item.host %}<a href="/hosts/{{ item.host.id }}">{{ item.host.name }}</a>{% else %}<span class="mono">{{ item.reg.host_id }}</span>{% endif %}</td>
+        <td>{% if item.reg.last_seen_at %}<span title="{{ item.reg.last_seen_at.strftime('%Y-%m-%d %H:%M:%S UTC') }}">{{ item.reg.last_seen_at|ago }}</span>{% else %}never{% endif %}</td>
+        <td>{{ item.reg.agent_version or "—" }}</td>
+        <td class="mono">{{ item.reg.host_ip or "—" }}</td>
+        <td>{{ item.reg.distro or "—" }}</td>
+        <td class="td-actions">
+          <form method="post" action="/plugins/host_agent/fleet/{{ item.reg.host_id }}/rotate-token"
+                style="display:inline;"
+                data-confirm="Rotate the token for {{ item.name }}? The existing agent will stop working until reinstalled with the new token.">{{ csrf_field() }}
+            <button type="submit" class="btn btn-ghost btn-sm">Rotate token</button>
+          </form>
+          <form method="post" action="/plugins/host_agent/fleet/{{ item.reg.host_id }}/delete"
+                style="display:inline;"
+                data-confirm="Unregister the agent for {{ item.name }}? Its token stops working and metrics stop until it is registered again. The agent itself stays installed on the host.">{{ csrf_field() }}
+            <button type="submit" class="btn btn-danger btn-sm">Unregister</button>
+          </form>
+        </td>
+      </tr>
+      {% else %}
+      <tr><td colspan="7" class="empty">No agents registered yet. <a href="/hosts/?add=1">Add a host</a>, then install its agent from the host's page.</td></tr>
+      {% endfor %}
+    </tbody>
+  </table>
+  </div>
 </div>
 
 {% if ansible_available %}
@@ -49,7 +74,7 @@
   <h3 style="margin-bottom:0.4rem;">Agent lifecycle via Ansible</h3>
   <p style="font-size:0.82rem;color:var(--text-muted);margin-bottom:0.6rem;">
     These actions run bundled <strong>Ansible playbooks</strong> to deploy and maintain the
-    Steward monitoring agent on your <a href="/ansible/inventory/targets">inventory hosts</a>.
+    Steward monitoring agent on your <a href="/ansible/inventory/">inventory hosts</a>.
     Steward generates each host's API token automatically and connects over SSH as the managed
     <code>steward</code> account — you'll be dropped into the live Ansible run to watch it.
   </p>
@@ -63,14 +88,14 @@
     <span style="font-size:0.84rem;flex:1;min-width:14rem;">
       No managed SSH key yet — Steward needs one to log into hosts as <code>steward</code>.
     </span>
-    <form method="post" action="/settings/ansible/generate-key" style="margin:0;">
+    <form method="post" action="/settings/ansible/generate-key" style="margin:0;">{{ csrf_field() }}
       <input type="hidden" name="next" value="/plugins/host_agent/fleet/">
       <button type="submit" class="btn btn-sm">Generate managed key</button>
     </form>
   </div>
   {% elif not (deploy_targets or deploy_groups) %}
   <p style="font-size:0.85rem;color:var(--text-dim);margin:0;">
-    No Ansible inventory targets yet. Add some under <a href="/ansible/inventory/targets">Ansible → Inventory</a>.
+    No Ansible inventory targets yet. Add some under <a href="/ansible/inventory/">Ansible → Inventory</a>.
   </p>
   {% endif %}
 </div>
@@ -85,7 +110,7 @@
     agent. Run this once per host.
   </p>
   <form method="post" action="/plugins/host_agent/provision"
-        style="display:flex;gap:0.75rem;align-items:flex-end;flex-wrap:wrap;">
+        style="display:flex;gap:0.75rem;align-items:flex-end;flex-wrap:wrap;">{{ csrf_field() }}
     <div class="form-group" style="margin-bottom:0;">{{ scope_select }}</div>
     <div class="form-group" style="margin-bottom:0;">
       <label>Bootstrap user</label>
@@ -111,7 +136,7 @@
     mints a fresh token, and installs the agent. No <code>curl | sh</code> needed.
   </p>
   <form method="post" action="/plugins/host_agent/deploy"
-        style="display:flex;gap:0.75rem;align-items:flex-end;flex-wrap:wrap;">
+        style="display:flex;gap:0.75rem;align-items:flex-end;flex-wrap:wrap;">{{ csrf_field() }}
     <div class="form-group" style="margin-bottom:0;">{{ scope_select }}</div>
     <div class="form-group" style="margin-bottom:0;">
       <label>Report interval (s)</label>
@@ -129,7 +154,7 @@
     so identity is preserved.
   </p>
   <form method="post" action="/plugins/host_agent/update"
-        style="display:flex;gap:0.75rem;align-items:flex-end;flex-wrap:wrap;">
+        style="display:flex;gap:0.75rem;align-items:flex-end;flex-wrap:wrap;">{{ csrf_field() }}
     <div class="form-group" style="margin-bottom:0;">{{ scope_select }}</div>
     <button type="submit" class="btn">Update</button>
   </form>
@@ -137,39 +162,4 @@
 {% endif %}
 {% endif %}
 
-<div class="card-flush">
-  <table class="table">
-    <thead>
-      <tr>
-        <th>Host</th><th>Reported IP</th><th>Agent version</th><th>Distro</th>
-        <th>Last seen</th><th class="td-actions">Actions</th>
-      </tr>
-    </thead>
-    <tbody>
-      {% for item in registrations %}
-      <tr>
-        <td>{{ item.host.name if item.host else item.reg.host_id }}</td>
-        <td>{{ item.reg.host_ip or "—" }}</td>
-        <td>{{ item.reg.agent_version or "—" }}</td>
-        <td>{{ item.reg.distro or "—" }}</td>
-        <td>{{ item.reg.last_seen_at.strftime("%Y-%m-%d %H:%M:%S") if item.reg.last_seen_at else "never" }}</td>
-        <td class="td-actions">
-          <form method="post" action="/plugins/host_agent/fleet/{{ item.reg.host_id }}/rotate-token"
-                style="display:inline;"
-                onsubmit="return confirm('Rotate the token for this host? The existing agent will stop working until reinstalled with the new token.');">
-            <button type="submit" class="btn btn-ghost btn-sm">Rotate token</button>
-          </form>
-          <form method="post" action="/plugins/host_agent/fleet/{{ item.reg.host_id }}/delete"
-                style="display:inline;"
-                onsubmit="return confirm('Delete this host registration? The agent will be unable to report metrics until re-registered.');">
-            <button type="submit" class="btn btn-danger btn-sm">Delete</button>
-          </form>
-        </td>
-      </tr>
-      {% else %}
-      <tr><td colspan="6" class="empty">No hosts registered. Add one above.</td></tr>
-      {% endfor %}
-    </tbody>
-  </table>
-</div>
 {% endblock %}

plugins/host_agent/templates/widget_history.html

@@ -16,16 +16,14 @@
      toggle that re-requests this widget without entering edit mode. #}
   <div style="flex:0 0 auto;display:flex;align-items:center;gap:0.5rem;flex-wrap:wrap;font-size:0.75rem;color:var(--text-muted);margin-bottom:0.3rem;">
     {% if session.user_id %}
-    <a href="/hosts/{{ host.id }}" style="font-weight:600;">{{ host.name }}</a>
+    <a href="/hosts/{{ host.id }}" style="font-weight:500;">{{ host.name }}</a>
     {% else %}
     <strong style="color:var(--text);">{{ host.name }}</strong>
     {% endif %}
-    <span class="hist-range" style="margin-left:auto;display:inline-flex;border:1px solid var(--border-mid);border-radius:5px;overflow:hidden;">
+    <span class="hist-range" style="margin-left:auto;display:inline-flex;gap:0.25rem;">
       {% for opt in [1, 6, 24] %}
-      <button type="button" onclick="histRange({{ wid }}, '{{ host.id }}', {{ opt }})"
-        style="border:0;cursor:pointer;padding:0.1rem 0.45rem;font-size:0.72rem;
-               background:{% if opt == hours %}var(--accent){% else %}transparent{% endif %};
-               color:{% if opt == hours %}#fff{% else %}var(--text-muted){% endif %};">{{ opt }}h</button>
+      <button type="button" onclick="histRange({{ wid|int }}, {{ opt|int }})"
+        class="range-btn{% if opt == hours %} active{% endif %}">{{ opt }}h</button>
       {% endfor %}
     </span>
   </div>
@@ -36,11 +34,15 @@
 <script>
 // Live time-range switch. Rewrites the parent cell's hx-get so the choice sticks
 // across polls (htmx re-reads the attribute each poll), then fetches immediately.
-window.histRange = function (wid, hostId, hours) {
+// Only `hours` changes: the rest of the query (host, widget id, a share link's
+// token) is kept, so the switch also works on a shared dashboard.
+window.histRange = function (wid, hours) {
   var cell = document.getElementById("widget-" + wid);
   if (!cell) return;
-  var base = (cell.getAttribute("hx-get") || "").split("?")[0];
-  var url = base + "?host_id=" + encodeURIComponent(hostId) + "&hours=" + hours + "&wid=" + wid;
+  var parts = (cell.getAttribute("hx-get") || "").split("?");
+  var query = new URLSearchParams(parts[1] || "");
+  query.set("hours", hours);
+  var url = parts[0] + "?" + query.toString();
   cell.setAttribute("hx-get", url);
   if (window.htmx) window.htmx.ajax("GET", url, { target: cell, swap: "innerHTML" });
 };
@@ -53,9 +55,9 @@ window.histRange = function (wid, hostId, hours) {
     return ("0" + d.getHours()).slice(-2) + ":" + ("0" + d.getMinutes()).slice(-2);
   };
   var ds = [
-    { key: "cpu_pct",      label: "CPU %",    color: "#c8a840" },
-    { key: "mem_used_pct", label: "Mem %",    color: "#4aa86a" },
-    { key: "disk_root",    label: "Disk / %", color: "#c87840" },
+    { key: "cpu_pct",      label: "CPU %",    color: stewardColor("--gold") },
+    { key: "mem_used_pct", label: "Mem %",    color: stewardColor("--green") },
+    { key: "disk_root",    label: "Disk / %", color: stewardColor("--yellow") },
   ];
   new Chart(el.getContext("2d"), {
     type: "line",
@@ -71,14 +73,14 @@ window.histRange = function (wid, hostId, hours) {
       interaction: { mode: "index", intersect: false },
       elements: { point: { radius: 0 } },
       scales: {
-        x: { type: "linear", ticks: { callback: function (v) { return fmtTime(v); }, maxTicksLimit: 6, color: "#8a8a92", font: { size: 10 } }, grid: { color: "#30303a" } },
+        x: { type: "linear", ticks: { callback: function (v) { return fmtTime(v); }, maxTicksLimit: 6, font: { size: 10 } } },
         // Fixed 0–{{ y_max }} ceiling (server snaps the data peak up to the next
         // 10% band). Steady across refreshes — auto-scaling made the zoom jump on
         // every poll — while a banded ceiling still keeps the lines filling the panel.
-        y: { min: 0, max: {{ y_max }}, ticks: { color: "#8a8a92", font: { size: 10 } }, grid: { color: "#30303a" } },
+        y: { min: 0, max: {{ y_max }}, ticks: { font: { size: 10 } } },
       },
       plugins: {
-        legend: { labels: { color: "#b8b8b0", boxWidth: 10, font: { size: 11 } } },
+        legend: { labels: { boxWidth: 10, font: { size: 11 } } },
         tooltip: { callbacks: { title: function (items) { return items.length ? fmtTime(items[0].parsed.x) : ""; } } },
       },
     },

plugins/host_agent/templates/widget_table.html

@@ -1,77 +0,0 @@
-{# host_agent widget — fleet glance. One horizontal row per host, zebra-striped
-   so adjacent hosts read as distinct groupings: health dot + host name on the
-   left, metric cells (cpu/mem/disk/load) each with a trend sparkline laid out
-   horizontally on the right. The name wraps (never hard-truncated) but the row
-   stays horizontal. #}
-{% if rows %}
-{% macro metric_cell(label, value, fmt, svg, style="", title="") %}
-<div style="min-width:62px;flex:0 0 auto;" title="{{ title }}">
-  <div style="font-size:0.62rem;color:var(--text-dim);text-transform:uppercase;letter-spacing:0.03em;">{{ label }}</div>
-  <div style="font-weight:600;font-size:0.8rem;font-variant-numeric:tabular-nums;{{ style }}">
-    {% if value is not none %}{{ fmt|format(value) }}{% else %}—{% endif %}
-  </div>
-  <div style="line-height:0;height:16px;">{{ svg | safe }}</div>
-</div>
-{% endmacro %}
-{# Human-readable throughput, matching the host-detail page. #}
-{% macro fmt_bps(v) %}
-{%- if v is none -%}—
-{%- elif v >= 1048576 -%}{{ "%.1f"|format(v / 1048576) }} MB/s
-{%- elif v >= 1024 -%}{{ "%.1f"|format(v / 1024) }} KB/s
-{%- else -%}{{ "%.0f"|format(v) }} B/s
-{%- endif -%}
-{% endmacro %}
-{# Two-line down/up (or read/write) throughput cell. #}
-{% macro io_cell(label, down_sym, down_val, up_sym, up_val, title="") %}
-<div style="min-width:80px;flex:0 0 auto;" title="{{ title }}">
-  <div style="font-size:0.62rem;color:var(--text-dim);text-transform:uppercase;letter-spacing:0.03em;">{{ label }}</div>
-  <div style="font-size:0.72rem;font-variant-numeric:tabular-nums;line-height:1.3;">
-    <div>{{ down_sym }} {{ fmt_bps(down_val) }}</div>
-    <div>{{ up_sym }} {{ fmt_bps(up_val) }}</div>
-  </div>
-</div>
-{% endmacro %}
-<div class="host-fleet">
-  {% for r in rows %}
-  {% set stale = r.stale %}
-  <div style="display:flex;align-items:center;gap:0.6rem 1rem;flex-wrap:wrap;padding:0.5rem 0.6rem;border-radius:4px;{% if loop.index0 % 2 %}background:var(--bg-elevated);{% endif %}">
-    {# Left — health dot + full host name (wraps, never truncated) + last seen #}
-    <div style="display:flex;align-items:center;gap:0.5rem;flex:1 1 150px;min-width:120px;">
-      <span class="dot {% if stale %}dot-dim{% elif (r.cpu_pct or 0) >= 90 or (r.mem_used_pct or 0) >= 90 or (r.disk_worst or 0) >= 90 %}dot-warn{% else %}dot-up{% endif %}"
-            style="flex-shrink:0;"
-            title="{% if stale %}stale / no recent data{% else %}warns at CPU/memory ≥90% or any disk mount ≥90% (the 'disk /' figure shows root only){% endif %}"></span>
-      <span style="min-width:0;">
-        {% if session.user_id %}
-        <a href="/hosts/{{ r.host.id }}" style="font-weight:600;font-size:0.85rem;overflow-wrap:anywhere;">{{ r.host.name }}</a>
-        {% else %}
-        <span style="font-weight:600;font-size:0.85rem;overflow-wrap:anywhere;">{{ r.host.name }}</span>
-        {% endif %}
-        <span style="display:block;font-size:0.7rem;color:var(--text-dim);">
-          {{ r.reg.last_seen_at.strftime("%H:%M:%S") if r.reg.last_seen_at else "never" }}
-        </span>
-      </span>
-    </div>
-    {# Right — metric cells with value + trend, laid out horizontally #}
-    <div style="display:flex;flex-wrap:wrap;gap:0.4rem 1rem;justify-content:flex-end;">
-      {{ metric_cell("cpu", r.cpu_pct, "%.0f%%", r.sparks.cpu, threshold_style(r.cpu_pct, 'cpu'), "Average CPU utilization") }}
-      {{ metric_cell("mem", r.mem_used_pct, "%.0f%%", r.sparks.mem, threshold_style(r.mem_used_pct, 'mem'), "Memory in use") }}
-      {{ metric_cell("disk /", r.disk_root, "%.0f%%", r.sparks.disk, threshold_style(r.disk_root, 'disk'), "Root filesystem (/) usage") }}
-      {{ metric_cell("load", r.load_1m, "%.2f", r.sparks.load, "", "Load average (1m)") }}
-      {# Extra agent metrics — shown only when the host reports them (VMs/containers
-         often have no temp sensor, etc.) so absent data doesn't clutter the row. #}
-      {% if r.net_rx_bps is not none or r.net_tx_bps is not none %}
-      {{ io_cell("net", "↓", r.net_rx_bps, "↑", r.net_tx_bps, "Network throughput (down / up)") }}
-      {% endif %}
-      {% if r.disk_read_bps is not none or r.disk_write_bps is not none %}
-      {{ io_cell("disk i/o", "rd", r.disk_read_bps, "wr", r.disk_write_bps, "Disk I/O (read / write)") }}
-      {% endif %}
-      {% if r.temp_max is not none %}
-      {{ metric_cell("temp", r.temp_max, "%.0f°C", "", threshold_style(r.temp_max, 'temp'), "Hottest sensor") }}
-      {% endif %}
-    </div>
-  </div>
-  {% endfor %}
-</div>
-{% else %}
-<p class="empty" style="padding:0.5rem 0;font-size:0.85rem;">No hosts with agent data yet.</p>
-{% endif %}

plugins/index.yaml

@@ -1,74 +0,0 @@
-# steward-plugins / index.yaml
-#
-# Plugin catalog for Steward.
-# Fetched by the app at: Settings → Plugins → Plugin Catalog
-#
-# After updating an entry, commit and push — changes are live within 5 minutes
-# (the app caches this file in-process for 300 seconds).
-#
-# See docs/plugins/index.yaml.example in the main app repo for the full schema.
-
-version: 1
-updated: "2026-03-22"
-
-plugins:
-
-  - name: docker
-    version: "2.0.0"
-    description: "Per-host Docker container status + resource usage, collected by the Steward host agent"
-    author: "Steward"
-    license: "MIT"
-    min_app_version: "0.1.0"
-    repository_url: "https://git.fabledsword.com/bvandeusen/Steward-plugins"
-    homepage: "https://git.fabledsword.com/bvandeusen/Steward-plugins/src/branch/main/docker"
-    download_url: "https://git.fabledsword.com/bvandeusen/Steward-plugins/releases/download/docker-v2.0.0/docker.zip"
-    checksum_sha256: ""
-    tags:
-      - containers
-      - docker
-      - infrastructure
-
-  - name: traefik
-    version: "1.0.0"
-    description: "Traefik reverse proxy metrics and access log integration"
-    author: "Steward"
-    license: "MIT"
-    min_app_version: "0.1.0"
-    repository_url: "https://git.fabledsword.com/bvandeusen/Steward-plugins"
-    homepage: "https://git.fabledsword.com/bvandeusen/Steward-plugins/src/branch/main/traefik"
-    download_url: "https://git.fabledsword.com/bvandeusen/Steward-plugins/releases/download/traefik-v1.0.0/traefik.zip"
-    checksum_sha256: ""
-    tags:
-      - proxy
-      - metrics
-      - access-log
-
-  - name: unifi
-    version: "1.0.0"
-    description: "UniFi Network controller integration — WAN health, devices, clients, DPI"
-    author: "Steward"
-    license: "MIT"
-    min_app_version: "0.1.0"
-    repository_url: "https://git.fabledsword.com/bvandeusen/Steward-plugins"
-    homepage: "https://git.fabledsword.com/bvandeusen/Steward-plugins/src/branch/main/unifi"
-    download_url: "https://git.fabledsword.com/bvandeusen/Steward-plugins/releases/download/unifi-v1.0.0/unifi.zip"
-    checksum_sha256: ""
-    tags:
-      - network
-      - unifi
-      - ubiquiti
-
-  - name: ups
-    version: "1.0.0"
-    description: "UPS monitoring via NUT (Network UPS Tools) with Ansible shutdown automation"
-    author: "Steward"
-    license: "MIT"
-    min_app_version: "0.1.0"
-    repository_url: "https://git.fabledsword.com/bvandeusen/Steward-plugins"
-    homepage: "https://git.fabledsword.com/bvandeusen/Steward-plugins/src/branch/main/ups"
-    download_url: "https://git.fabledsword.com/bvandeusen/Steward-plugins/releases/download/ups-v1.0.0/ups.zip"
-    checksum_sha256: ""
-    tags:
-      - ups
-      - power
-      - nut

plugins/prometheus/__init__.py

@@ -0,0 +1,70 @@
+# plugins/prometheus/__init__.py
+from __future__ import annotations
+from typing import TYPE_CHECKING
+
+if TYPE_CHECKING:
+    from quart import Quart
+
+_app: "Quart | None" = None
+
+
+def setup(app: "Quart") -> None:
+    global _app
+    _app = app
+    from .models import PrometheusTarget, PrometheusWidgetTemplate  # noqa: F401 — registers the tables
+    register_widgets()
+
+
+def register_widgets() -> None:
+    """One dashboard widget for every target and template (System 181's
+    ruling). Its target and template lists are read when the widget picker
+    opens, so a new target or template appears without a restart."""
+    from steward.core.widgets import register_widget
+
+    from . import library
+    register_widget({
+        "key": "prometheus",
+        "label": "Prometheus",
+        "description": "One scrape target drawn by a widget template: Fabled Archives, or one you copy in",
+        "hx_url": "/plugins/prometheus/widget",
+        "detail_url": "/plugins/prometheus/targets/{target}",
+        "plugin": "prometheus",
+        "poll": True,
+        "default_title": library.default_title,
+        "params": [
+            {"key": "target", "label": "Target", "type": "select", "default": "",
+             "options_from": library.target_options,
+             "empty": "No targets yet: add one under Prometheus first."},
+            {"key": "template", "label": "Template", "type": "select", "default": "",
+             "options_from": library.template_options,
+             "empty": "No templates yet."},
+        ],
+    })
+
+
+def get_scheduled_tasks() -> list:
+    from .scheduler import make_scrape_task
+    return [make_scrape_task(_app)]
+
+
+def get_blueprint():
+    from .routes import prometheus_bp
+    return prometheus_bp
+
+
+def get_nav() -> list[dict]:
+    # Surfaces in the sidebar's Infrastructure group (scraped endpoints).
+    return [{"label": "Prometheus", "href": "/plugins/prometheus/"}]
+
+
+def get_alert_metrics() -> dict[str, list[str]]:
+    """The metric names are whatever each endpoint exposes, so the list is
+    empty here and the rule form offers the names recorded so far."""
+    return {"prometheus": []}
+
+
+def get_audit_entities() -> dict:
+    """Audit rows about a target or a widget template name it and link to its page."""
+    from .models import PrometheusTarget, PrometheusWidgetTemplate
+    return {"prometheus_target": (PrometheusTarget, "/plugins/prometheus/targets/{}"),
+            "prometheus_template": (PrometheusWidgetTemplate, "/plugins/prometheus/templates/custom:{}")}

plugins/prometheus/library.py

@@ -0,0 +1,140 @@
+"""The widget templates a Prometheus widget can draw with.
+
+Bundled templates ship in plugins/prometheus/widgets/*.yaml, read once and
+read-only; their key is "bundled:<file name>". Templates an admin writes or
+copies in are rows (PrometheusWidgetTemplate, #5692); their key is
+"custom:<id>". Steward fetches none from anywhere (System 181's ruling).
+"""
+from __future__ import annotations
+
+import json
+import re
+from dataclasses import dataclass, field
+from datetime import datetime
+from functools import lru_cache
+from pathlib import Path
+
+from sqlalchemy import select
+
+from .models import PrometheusTarget, PrometheusWidgetTemplate
+from .template_format import Problem, Template, TemplateError, check
+
+BUNDLED = Path(__file__).parent / "widgets"
+CUSTOM = "custom:"
+
+
+@dataclass
+class Entry:
+    key: str
+    name: str
+    bundled: bool
+    text: str
+    template: Template | None = None
+    problems: list[Problem] = field(default_factory=list)
+    id: str | None = None                   # a custom template's row
+    updated_at: datetime | None = None
+
+
+def _entry(key: str, fallback_name: str, text: str, bundled: bool, **row) -> Entry:
+    try:
+        template = check(text)
+    except TemplateError as exc:
+        return Entry(key, fallback_name, bundled, text, None, exc.problems, **row)
+    return Entry(key, template.name, bundled, text, template, **row)
+
+
+@lru_cache(maxsize=1)
+def bundled() -> tuple[Entry, ...]:
+    return tuple(_entry(f"bundled:{p.stem}", p.stem, p.read_text(encoding="utf-8"), True)
+                 for p in sorted(BUNDLED.glob("*.yaml")))
+
+
+def custom_entry(row: PrometheusWidgetTemplate) -> Entry:
+    # Saved only once it passed, but a later Steward may read the format
+    # differently: it is checked again, and says so if it no longer passes.
+    return _entry(CUSTOM + row.id, row.name, row.text, False, id=row.id, updated_at=row.updated_at)
+
+
+async def entries(db) -> list[Entry]:
+    """Every template, by name."""
+    rows = (await db.execute(select(PrometheusWidgetTemplate))).scalars().all()
+    return sorted([*bundled(), *(custom_entry(r) for r in rows)], key=lambda e: e.name.lower())
+
+
+async def find(db, key: str) -> Entry | None:
+    if key.startswith(CUSTOM):
+        row = await db.get(PrometheusWidgetTemplate, key[len(CUSTOM):])
+        return custom_entry(row) if row else None
+    return next((e for e in bundled() if e.key == key), None)
+
+
+async def name_taken(db, name: str, other_than: str | None = None) -> bool:
+    """Another saved template has this name; the widget picker lists by name."""
+    q = select(PrometheusWidgetTemplate.id).where(PrometheusWidgetTemplate.name == name)
+    if other_than:
+        q = q.where(PrometheusWidgetTemplate.id != other_than)
+    return (await db.execute(q)).first() is not None
+
+
+async def used_by(db) -> dict[str, int]:
+    """How many dashboard widgets draw with each template key."""
+    from steward.models.dashboard import DashboardWidget
+    counts: dict[str, int] = {}
+    rows = (await db.execute(select(DashboardWidget.config_json)
+                             .where(DashboardWidget.widget_key == "prometheus"))).scalars()
+    for raw in rows:
+        try:
+            key = (json.loads(raw or "{}") or {}).get("template")
+        except (ValueError, AttributeError):
+            continue
+        if key:
+            counts[key] = counts.get(key, 0) + 1
+    return counts
+
+
+# ── Copying ──────────────────────────────────────────────────────────────────
+
+_NAME_LINE = re.compile(r"^name:[^\n]*$", re.MULTILINE)
+
+
+def renamed(text: str, name: str) -> str:
+    """`text` with its top-level `name:` set to `name`, the rest as written
+    (its comments and layout kept)."""
+    line = "name: " + json.dumps(name, ensure_ascii=False)
+    if _NAME_LINE.search(text):
+        return _NAME_LINE.sub(lambda _: line, text, count=1)
+    return f"{line}\n{text}"
+
+
+def copy_name(name: str, taken: set[str]) -> str:
+    """"Fabled Archives (copy)", then "(copy 2)", … a name nobody has."""
+    for n in range(1, 1000):
+        candidate = f"{name} (copy)" if n == 1 else f"{name} (copy {n})"
+        if candidate not in taken:
+            return candidate[:80]
+    return name[:80]
+
+
+def file_name(name: str) -> str:
+    """A download's file name: "Fabled Archives" → fabled-archives.yaml."""
+    slug = re.sub(r"[^a-z0-9]+", "-", name.lower()).strip("-")
+    return f"{slug or 'template'}.yaml"
+
+
+# ── The widget's settings ────────────────────────────────────────────────────
+
+async def target_options(db) -> list[dict]:
+    rows = (await db.execute(select(PrometheusTarget.id, PrometheusTarget.name)
+                             .order_by(PrometheusTarget.name))).all()
+    return [{"value": tid, "label": name} for tid, name in rows]
+
+
+async def template_options(db) -> list[dict]:
+    return [{"value": e.key, "label": f"{e.name} (built in)" if e.bundled else e.name}
+            for e in await entries(db) if e.template is not None]
+
+
+async def default_title(db, config: dict) -> str | None:
+    """A new widget is named after its template: "Fabled Archives"."""
+    entry = await find(db, str(config.get("template") or ""))
+    return entry.name if entry and entry.template else None

plugins/prometheus/migrations/versions/prometheus_001_targets.py

@@ -0,0 +1,46 @@
+"""Prometheus scrape targets are rows edited on the Prometheus page
+
+Each row is a /metrics endpoint: where it is, its bearer token (encrypted),
+how often to scrape it, which series to keep, and the newest scrape's
+outcome (milestone 506).
+
+Revision ID: prometheus_001_targets
+Revises: (none — a branch off core via depends_on)
+Create Date: 2026-10-12
+"""
+from typing import Sequence, Union
+
+import sqlalchemy as sa
+from alembic import op
+
+revision: str = "prometheus_001_targets"
+down_revision: Union[str, None] = None
+branch_labels: Union[str, Sequence[str], None] = "prometheus"
+depends_on: Union[str, Sequence[str], None] = ("0004_app_settings",)
+
+
+def upgrade() -> None:
+    op.create_table(
+        "prometheus_targets",
+        sa.Column("id", sa.String(36), primary_key=True),
+        sa.Column("name", sa.String(100), nullable=False, unique=True),
+        sa.Column("url", sa.String(500), nullable=False),
+        sa.Column("token", sa.Text, nullable=False, server_default=""),
+        sa.Column("verify_tls", sa.Boolean, nullable=False, server_default=sa.true()),
+        sa.Column("interval_seconds", sa.Integer, nullable=False, server_default="60"),
+        sa.Column("keep", sa.JSON, nullable=False, server_default="[]"),
+        sa.Column("drop", sa.JSON, nullable=False, server_default="[]"),
+        sa.Column("series_cap", sa.Integer, nullable=False, server_default="1000"),
+        sa.Column("enabled", sa.Boolean, nullable=False, server_default=sa.true()),
+        sa.Column("created_at", sa.DateTime(timezone=True), nullable=False),
+        sa.Column("last_scraped_at", sa.DateTime(timezone=True), nullable=True),
+        sa.Column("last_ok_at", sa.DateTime(timezone=True), nullable=True),
+        sa.Column("last_error", sa.Text, nullable=True),
+        sa.Column("last_kept", sa.Integer, nullable=False, server_default="0"),
+        sa.Column("last_dropped", sa.Integer, nullable=False, server_default="0"),
+        sa.Column("last_capped", sa.Integer, nullable=False, server_default="0"),
+    )
+
+
+def downgrade() -> None:
+    op.drop_table("prometheus_targets")

plugins/prometheus/migrations/versions/prometheus_002_templates.py

@@ -0,0 +1,33 @@
+"""Widget templates an admin writes or copies in are rows
+
+Each row is one template's YAML and its name, beside the bundled templates
+that ship as files (#5692).
+
+Revision ID: prometheus_002_templates
+Revises: prometheus_001_targets
+Create Date: 2026-10-14
+"""
+from typing import Sequence, Union
+
+import sqlalchemy as sa
+from alembic import op
+
+revision: str = "prometheus_002_templates"
+down_revision: Union[str, None] = "prometheus_001_targets"
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    op.create_table(
+        "prometheus_widget_templates",
+        sa.Column("id", sa.String(36), primary_key=True),
+        sa.Column("name", sa.String(80), nullable=False, unique=True),
+        sa.Column("text", sa.Text, nullable=False),
+        sa.Column("created_at", sa.DateTime(timezone=True), nullable=False),
+        sa.Column("updated_at", sa.DateTime(timezone=True), nullable=False),
+    )
+
+
+def downgrade() -> None:
+    op.drop_table("prometheus_widget_templates")

plugins/prometheus/models.py

@@ -0,0 +1,59 @@
+# plugins/prometheus/models.py
+from __future__ import annotations
+
+from datetime import datetime, timezone
+
+from sqlalchemy import JSON, Boolean, DateTime, Integer, String, Text
+from sqlalchemy.orm import Mapped, mapped_column
+
+from steward.core.ids import time_ordered_id
+from steward.models.base import Base
+
+
+class PrometheusTarget(Base):
+    """A /metrics endpoint Steward scrapes, edited on the Prometheus page.
+
+    The bearer token is stored encrypted. keep and drop are lists of name
+    patterns ("fabledarchives_*") an admin chooses: Steward does the
+    filtering, the endpoint serves everything. The last_* columns are the
+    newest scrape's outcome, written by the scraper, so a page can say why a
+    target isn't answering without asking it.
+    """
+    __tablename__ = "prometheus_targets"
+
+    id: Mapped[str] = mapped_column(String(36), primary_key=True, default=time_ordered_id)
+    name: Mapped[str] = mapped_column(String(100), nullable=False, unique=True)
+    url: Mapped[str] = mapped_column(String(500), nullable=False)
+    token: Mapped[str] = mapped_column(Text, nullable=False, default="")
+    verify_tls: Mapped[bool] = mapped_column(Boolean, nullable=False, default=True)
+    interval_seconds: Mapped[int] = mapped_column(Integer, nullable=False, default=60)
+    keep: Mapped[list] = mapped_column(JSON, nullable=False, default=list)
+    drop: Mapped[list] = mapped_column(JSON, nullable=False, default=list)
+    series_cap: Mapped[int] = mapped_column(Integer, nullable=False, default=1000)
+    enabled: Mapped[bool] = mapped_column(Boolean, nullable=False, default=True)
+    created_at: Mapped[datetime] = mapped_column(
+        DateTime(timezone=True), nullable=False, default=lambda: datetime.now(timezone.utc))
+    last_scraped_at: Mapped[datetime | None] = mapped_column(DateTime(timezone=True), nullable=True)
+    last_ok_at: Mapped[datetime | None] = mapped_column(DateTime(timezone=True), nullable=True)
+    last_error: Mapped[str | None] = mapped_column(Text, nullable=True)
+    last_kept: Mapped[int] = mapped_column(Integer, nullable=False, default=0)
+    last_dropped: Mapped[int] = mapped_column(Integer, nullable=False, default=0)
+    last_capped: Mapped[int] = mapped_column(Integer, nullable=False, default=0)
+
+
+class PrometheusWidgetTemplate(Base):
+    """A widget template an admin wrote or copied in (#5692), beside the
+    bundled ones in widgets/*.yaml. `text` is the YAML as written; `name` is
+    its `name:`, kept here so the library lists and orders without reading
+    every template. Only a template that passes the checker is saved. Its key
+    in a widget's settings is "custom:<id>", so renaming it keeps its widgets.
+    """
+    __tablename__ = "prometheus_widget_templates"
+
+    id: Mapped[str] = mapped_column(String(36), primary_key=True, default=time_ordered_id)
+    name: Mapped[str] = mapped_column(String(80), nullable=False, unique=True)
+    text: Mapped[str] = mapped_column(Text, nullable=False)
+    created_at: Mapped[datetime] = mapped_column(
+        DateTime(timezone=True), nullable=False, default=lambda: datetime.now(timezone.utc))
+    updated_at: Mapped[datetime] = mapped_column(
+        DateTime(timezone=True), nullable=False, default=lambda: datetime.now(timezone.utc))

plugins/prometheus/plugin.yaml

@@ -0,0 +1,17 @@
+# plugins/prometheus/plugin.yaml
+name: prometheus
+version: "1.0.0"
+description: "Scrapes Prometheus /metrics endpoints into Steward's metrics, so any service that exposes them can be charted and alerted on."
+author: "Steward"
+license: "MIT"
+min_app_version: "0.1.0"
+repository_url: "https://git.fabledsword.com/bvandeusen/FabledSteward"
+homepage: "https://git.fabledsword.com/bvandeusen/FabledSteward/src/branch/dev/plugins/prometheus"
+tags:
+  - prometheus
+  - metrics
+  - monitoring
+
+# Targets are added, tested and edited on the Prometheus page
+# (/plugins/prometheus/); each carries its own interval.
+config: {}

plugins/prometheus/routes.py

@@ -0,0 +1,582 @@
+"""The Prometheus pages. Every GET reads Steward's own records: the target
+rows the editor writes and the scraper annotates, and the series it stores.
+Only Test, an admin's explicit action, asks an endpoint anything (System
+178's ruling)."""
+from __future__ import annotations
+
+import json
+from datetime import datetime, timedelta, timezone
+
+from quart import Blueprint, Response, abort, current_app, redirect, render_template, request, session
+from sqlalchemy import delete, func, select, update
+
+from steward.auth.middleware import require_role
+from steward.core.audit import log_audit
+from steward.core.crypto import decrypt_secret
+from steward.core.metric_series import latest_values, points
+from steward.models.metrics import PluginMetric, PluginMetricLatest
+from steward.models.users import UserRole
+
+from . import library
+from .models import PrometheusTarget, PrometheusWidgetTemplate
+from .scraper import ScrapeError, scrape
+from .service import (OWN_METRICS, SOURCE, FormError, fold_states, form_values, map_families,
+                      parse_form, patterns, shown_value, target_state)
+
+prometheus_bp = Blueprint("prometheus", __name__, template_folder="templates")
+
+_REACH_ORDER = {"down": 0, "stale": 1, "pending": 2, "up": 3, "off": 4}
+
+
+def _actor() -> tuple[str | None, str]:
+    return session.get("user_id"), session.get("username", "")
+
+
+async def _all_targets(db) -> list[PrometheusTarget]:
+    return list((await db.execute(select(PrometheusTarget).order_by(PrometheusTarget.name))).scalars().all())
+
+
+def view(target: PrometheusTarget, own: dict, now: datetime) -> dict:
+    """What a page says about a target: its state in words and the numbers
+    from its newest scrape."""
+    state = target_state(target, now)
+    return {"id": target.id, "name": target.name, "url": target.url, "reach": state.reach,
+            "says": state.says, "scraped_at": target.last_scraped_at, "ok_at": target.last_ok_at,
+            "interval": target.interval_seconds, "kept": target.last_kept,
+            "dropped": target.last_dropped, "capped": target.last_capped, "cap": target.series_cap,
+            "keep": list(target.keep or []), "drop": list(target.drop or []),
+            "scrape_ms": own.get("scrape_ms")}
+
+
+@prometheus_bp.get("/")
+@require_role(UserRole.viewer)
+async def index():
+    async with current_app.db_sessionmaker() as db:
+        targets = await _all_targets(db)
+        own = {}
+        if targets:
+            rows = (await db.execute(select(PluginMetricLatest).where(
+                PluginMetricLatest.source_module == SOURCE,
+                PluginMetricLatest.resource_name.in_([t.name for t in targets]),
+                PluginMetricLatest.metric_name == "scrape_ms"))).scalars()
+            own = {r.resource_name: {"scrape_ms": r.value} for r in rows}
+    now = datetime.now(timezone.utc)
+    shown = [view(t, own.get(t.name, {}), now) for t in targets]
+    shown.sort(key=lambda v: (_REACH_ORDER[v["reach"]], v["name"].lower()))
+    return await render_template("prometheus/index.html", targets=shown)
+
+
+# ── The editor ────────────────────────────────────────────────────────────────
+
+async def _render_form(target: PrometheusTarget | None, values: dict, error: str | None = None,
+                       status: int = 200):
+    return await render_template("prometheus/form.html", target=target, v=values, error=error), status
+
+
+def _typed(form, target: PrometheusTarget | None = None) -> dict:
+    """What was typed, to show the form again with an error. The token is not
+    echoed back."""
+    return {"name": form.get("name", ""), "url": form.get("url", ""),
+            "interval_seconds": form.get("interval_seconds", ""), "series_cap": form.get("series_cap", ""),
+            "keep": form.get("keep", ""), "drop": form.get("drop", ""),
+            "verify_tls": bool(form.get("verify_tls")), "enabled": bool(form.get("enabled")),
+            "has_token": bool(target and target.token)}
+
+
+async def _name_taken(db, name: str, other_than: str | None = None) -> bool:
+    q = select(PrometheusTarget.id).where(PrometheusTarget.name == name)
+    if other_than:
+        q = q.where(PrometheusTarget.id != other_than)
+    return (await db.execute(q)).first() is not None
+
+
+async def _target_or_404(db, target_id: str) -> PrometheusTarget:
+    target = await db.get(PrometheusTarget, target_id)
+    if target is None:
+        abort(404)
+    return target
+
+
+@prometheus_bp.get("/targets/new")
+@require_role(UserRole.admin)
+async def new_target():
+    return await _render_form(None, form_values(None))
+
+
+@prometheus_bp.post("/targets/new")
+@require_role(UserRole.admin)
+async def create_target():
+    form = await request.form
+    try:
+        values = parse_form(form)
+    except FormError as exc:
+        return await _render_form(None, _typed(form), str(exc), 422)
+    async with current_app.db_sessionmaker() as db:
+        async with db.begin():
+            if await _name_taken(db, values["name"]):
+                return await _render_form(None, _typed(form),
+                                          f"A target is already named “{values['name']}”.", 422)
+            target = PrometheusTarget(**values)
+            db.add(target)
+            await db.flush()
+            target_id = target.id
+    await log_audit(current_app, *_actor(), "prometheus.target_created", entity_type="prometheus_target",
+                    entity_id=target_id, detail={"name": values["name"], "url": values["url"]})
+    return redirect(f"/plugins/prometheus/targets/{target_id}")
+
+
+@prometheus_bp.get("/targets/<target_id>/edit")
+@require_role(UserRole.admin)
+async def edit_target(target_id: str):
+    async with current_app.db_sessionmaker() as db:
+        target = await _target_or_404(db, target_id)
+    return await _render_form(target, form_values(target))
+
+
+@prometheus_bp.post("/targets/<target_id>/edit")
+@require_role(UserRole.admin)
+async def save_target(target_id: str):
+    form = await request.form
+    async with current_app.db_sessionmaker() as db:
+        async with db.begin():
+            target = await _target_or_404(db, target_id)
+            try:
+                values = parse_form(form, target)
+            except FormError as exc:
+                return await _render_form(target, _typed(form, target), str(exc), 422)
+            if await _name_taken(db, values["name"], other_than=target.id):
+                return await _render_form(target, _typed(form, target),
+                                          f"A target is already named “{values['name']}”.", 422)
+            old_name = target.name
+            for column, value in values.items():
+                setattr(target, column, value)
+            # Where it is or what is kept may have changed: the last outcome no
+            # longer describes it, so the page says "not scraped yet" until the next scrape.
+            target.last_scraped_at = target.last_ok_at = target.last_error = None
+            target.last_kept = target.last_dropped = target.last_capped = 0
+            if old_name != target.name:
+                # Its history is stored under its name, and its labelled series
+                # under "<name>:…": move them with the rename.
+                for model in (PluginMetric, PluginMetricLatest):
+                    await db.execute(update(model).where(
+                        model.source_module == SOURCE, model.resource_name == old_name
+                    ).values(resource_name=target.name))
+                    # "<old>:labels" keeps its ":labels" (substr is 1-based).
+                    await db.execute(update(model).where(
+                        model.source_module == SOURCE,
+                        model.resource_name.startswith(old_name + ":", autoescape=True),
+                    ).values(resource_name=target.name + func.substr(model.resource_name, len(old_name) + 1)))
+    from . import scheduler
+    scheduler.RATES.forget(target_id)
+    detail = {"name": values["name"]}
+    if old_name != values["name"]:
+        detail["old_name"] = old_name
+    await log_audit(current_app, *_actor(), "prometheus.target_updated", entity_type="prometheus_target",
+                    entity_id=target_id, detail=detail)
+    return redirect(f"/plugins/prometheus/targets/{target_id}")
+
+
+@prometheus_bp.post("/targets/<target_id>/delete")
+@require_role(UserRole.admin)
+async def delete_target(target_id: str):
+    async with current_app.db_sessionmaker() as db:
+        async with db.begin():
+            target = await _target_or_404(db, target_id)
+            name = target.name
+            await db.delete(target)
+            # Its newest values go with it, so nothing reads it as a target
+            # that stopped answering. Its history ages out with retention.
+            await db.execute(delete(PluginMetricLatest).where(
+                PluginMetricLatest.source_module == SOURCE,
+                (PluginMetricLatest.resource_name == name)
+                | PluginMetricLatest.resource_name.startswith(name + ":", autoescape=True)))
+    from . import scheduler
+    scheduler.RATES.forget(target_id)
+    await log_audit(current_app, *_actor(), "prometheus.target_deleted", entity_type="prometheus_target",
+                    entity_id=target_id, detail={"name": name})
+    return redirect("/plugins/prometheus/")
+
+
+TEST_FAMILIES_SHOWN = 40
+
+
+@prometheus_bp.post("/targets/test")
+@require_role(UserRole.admin)
+async def test_target():
+    """Scrape the (unsaved) form's address once and say what would be stored.
+    An admin's explicit action, so it may contact the endpoint."""
+    form = await request.form
+    existing = None
+    if form.get("target_id"):
+        async with current_app.db_sessionmaker() as db:
+            existing = await db.get(PrometheusTarget, form["target_id"])
+    url = (form.get("url") or "").strip()
+    token = (form.get("token") or "").strip()
+    if not token and existing is not None and existing.token and not form.get("clear_token"):
+        token = decrypt_secret(existing.token, context=f"prometheus target {existing.name} token")
+    found: dict = {}
+    if not url.startswith(("http://", "https://")):
+        found["error"] = "Enter the address first: http://… or https://…"
+    else:
+        try:
+            families = await scrape(url, token, bool(form.get("verify_tls")))
+        except ScrapeError as exc:
+            found["error"] = str(exc)
+        else:
+            try:
+                cap = max(1, int(form.get("series_cap") or 1000))
+            except ValueError:
+                cap = 1000
+            keep, drop = patterns(form.get("keep") or ""), patterns(form.get("drop") or "")
+            mapped = map_families("t", families, keep, drop, cap)
+            kept_names = {}
+            for s in mapped.series:
+                kept_names[s.metric] = kept_names.get(s.metric, 0) + 1
+            found = {"families": len(families), "kept": len(mapped.series), "dropped": mapped.dropped,
+                     "capped": mapped.capped,
+                     "names": sorted(kept_names.items())[:TEST_FAMILIES_SHOWN],
+                     "more": max(0, len(kept_names) - TEST_FAMILIES_SHOWN)}
+    return await render_template("prometheus/_test.html", found=found, url=url)
+
+
+# ── The target page and one series ───────────────────────────────────────────
+
+def group_series(target_name: str, latest: list) -> list[dict]:
+    """The target's stored series grouped by metric, each with its rows:
+    the labels that tell them apart, and the newest value. A state set is
+    folded to one row per thing it describes, naming the current state
+    (`states` set on the group)."""
+    groups: dict[str, list] = {}
+    for row in latest:
+        if row.resource_name == target_name and row.metric_name in OWN_METRICS:
+            continue
+        labels = row.resource_name[len(target_name) + 1:] if row.resource_name != target_name else ""
+        groups.setdefault(row.metric_name, []).append({
+            "labels": labels, "resource": row.resource_name, "at": row.recorded_at,
+            "value": row.value, **shown_value(row.metric_name, row.value)})
+    out = []
+    for metric, rows in sorted(groups.items()):
+        folded = fold_states(metric, rows)
+        if folded is not None:
+            out.append({"metric": metric, "states": True, "rows": [
+                {"labels": f["labels"], "resource": f["resource"], "moment": None,
+                 "text": f["state"] or "none", "states": f["states"]} for f in folded]})
+        else:
+            out.append({"metric": metric, "states": False,
+                        "rows": sorted(rows, key=lambda r: r["labels"])})
+    return out
+
+
+@prometheus_bp.get("/targets/<target_id>")
+@require_role(UserRole.viewer)
+async def target_detail(target_id: str):
+    q = (request.args.get("q") or "").strip()
+    async with current_app.db_sessionmaker() as db:
+        target = await _target_or_404(db, target_id)
+        latest = await latest_values(db, SOURCE, target.name, include_children=True)
+    own = {r.metric_name: r.value for r in latest if r.resource_name == target.name
+           and r.metric_name in OWN_METRICS}
+    groups = group_series(target.name, latest)
+    total = sum(len(g["rows"]) for g in groups)
+    if q:
+        groups = [g for g in groups if q.lower() in g["metric"].lower()]
+    now = datetime.now(timezone.utc)
+    return await render_template("prometheus/target.html", target=view(target, own, now), groups=groups,
+                                 total=total, q=q)
+
+
+RANGES = (1, 6, 24, 168)
+
+
+@prometheus_bp.get("/targets/<target_id>/series")
+@require_role(UserRole.viewer)
+async def series_detail(target_id: str):
+    resource = request.args.get("resource") or ""
+    metric = request.args.get("metric") or ""
+    hours = request.args.get("hours", type=int, default=24)
+    if hours not in RANGES:
+        hours = 24
+    async with current_app.db_sessionmaker() as db:
+        target = await _target_or_404(db, target_id)
+        if resource != target.name and not resource.startswith(target.name + ":"):
+            abort(404)
+        now_row = await db.get(PluginMetricLatest, (SOURCE, resource, metric))
+        series = await points(db, SOURCE, resource, metric,
+                              datetime.now(timezone.utc) - timedelta(hours=hours))
+    labels = resource[len(target.name) + 1:] if resource != target.name else ""
+    stamp_axis = metric.endswith("_timestamp_seconds")
+    return await render_template(
+        "prometheus/series.html", target=target, resource=resource, metric=metric, labels=labels,
+        now=shown_value(metric, now_row.value if now_row else None), hours=hours, ranges=RANGES,
+        points_json=json.dumps([(t.isoformat(), v) for t, v in series]), stamp_axis=stamp_axis)
+
+
+# ── The dashboard widget ─────────────────────────────────────────────────────
+
+SPARK_HOURS = 24
+
+
+def spark_heights(values: list[float]) -> list[float]:
+    """Bar heights in % of the series' own peak over the window (System
+    154's ruling on small history bars); a flat zero draws flat."""
+    peak = max((abs(v) for v in values), default=0.0)
+    return [round(100.0 * abs(v) / peak, 1) if peak else 0.0 for v in values]
+
+
+async def draw(db, template, target: PrometheusTarget, now: datetime):
+    """`template` drawn over `target`'s stored values, with the last day's bars
+    for the items that ask for them. The dashboard widget and the template
+    editor's preview both draw through here."""
+    from .template_view import render
+
+    latest = await latest_values(db, SOURCE, target.name, include_children=True)
+    shown = render(template, target.name, latest, now)
+    if shown.wants:
+        since = now - timedelta(hours=SPARK_HOURS)
+        history = {}
+        for resource, metric in shown.wants:
+            got = await points(db, SOURCE, resource, metric, since, target=SPARK_HOURS)
+            history[(resource, metric)] = spark_heights([v for _, v in got])
+        shown = render(template, target.name, latest, now, history=history)
+    return shown
+
+
+@prometheus_bp.get("/widget")
+@require_role(UserRole.viewer)
+async def widget():
+    """The Prometheus dashboard widget: one target drawn by one template
+    (System 181's ruling), in widget shape. Reads stored values only."""
+
+    target_id = request.args.get("target") or ""
+    key = request.args.get("template") or ""
+    now = datetime.now(timezone.utc)
+    async with current_app.db_sessionmaker() as db:
+        target = await db.get(PrometheusTarget, target_id) if target_id else None
+        entry = await library.find(db, key) if key else None
+        problem = ("This widget has no target; remove it and add it again with one." if not target_id
+                   else "The target this widget showed has been deleted." if target is None
+                   else "This widget has no template; remove it and add it again with one." if not key
+                   else "The template this widget used is gone." if entry is None
+                   else f"The template “{entry.name}” has a problem: {entry.problems[0]}"
+                   if entry.template is None else "")
+        if problem:
+            return await render_template("prometheus/widget.html", problem=problem, target=None,
+                                         view=None)
+        shown = await draw(db, entry.template, target, now)
+    return await render_template("prometheus/widget.html", problem="", view=shown,
+                                 target=view_target(target, now), spark_hours=SPARK_HOURS)
+
+
+def view_target(target: PrometheusTarget, now: datetime) -> dict:
+    state = target_state(target, now)
+    return {"id": target.id, "name": target.name, "reach": state.reach, "says": state.says,
+            "ok_at": target.last_ok_at, "scraped_at": target.last_scraped_at}
+
+
+# ── The template library (#5692) ─────────────────────────────────────────────
+# Bundled templates are read-only: they can be looked at, exported and
+# duplicated. An admin writes, pastes or uploads their own, edits and deletes
+# them. Nothing is fetched from anywhere (System 181's ruling).
+
+def _shown_entry(entry, used: int) -> dict:
+    return {"key": entry.key, "id": entry.id, "name": entry.name, "bundled": entry.bundled,
+            "version": entry.template.version if entry.template else None,
+            "description": entry.template.description if entry.template else "",
+            "problem": str(entry.problems[0]) if entry.problems else "",
+            "updated_at": entry.updated_at, "used": used}
+
+
+@prometheus_bp.get("/templates/")
+@require_role(UserRole.viewer)
+async def templates_index():
+    async with current_app.db_sessionmaker() as db:
+        entries = await library.entries(db)
+        used = await library.used_by(db)
+    shown = [_shown_entry(e, used.get(e.key, 0)) for e in entries]
+    # One that no longer passes the checker first: its widgets say so too.
+    shown.sort(key=lambda t: (not t["problem"], t["name"].lower()))
+    return await render_template("prometheus/templates.html", templates=shown)
+
+
+async def _editor(entry=None, text: str = "", problems=(), note: str = "", status: int = 200,
+                  saving: bool = False):
+    """The template page: an admin's editor, or a read-only view of a
+    bundled template (or of any template, to a viewer). Both preview it."""
+    async with current_app.db_sessionmaker() as db:
+        targets = await library.target_options(db)
+        used = (await library.used_by(db)).get(entry.key, 0) if entry is not None else 0
+    editable = session.get("user_role") == "admin" and (entry is None or not entry.bundled)
+    return await render_template(
+        "prometheus/template.html", entry=entry, title=entry.name if entry else "New template",
+        text=text if entry is None or text else entry.text,
+        problems=[str(p) for p in problems], note=note, targets=targets, editable=editable,
+        used=used, saving=saving), status
+
+
+async def _entry_or_404(db, key: str):
+    entry = await library.find(db, key)
+    if entry is None:
+        abort(404)
+    return entry
+
+
+@prometheus_bp.get("/templates/new")
+@require_role(UserRole.admin)
+async def new_template():
+    """A blank editor, or a copy of another template (?from=<key>) under a
+    name nobody has."""
+    source = request.args.get("from") or ""
+    if not source:
+        return await _editor()
+    async with current_app.db_sessionmaker() as db:
+        entry = await _entry_or_404(db, source)
+        taken = {e.name for e in await library.entries(db)}
+    return await _editor(text=library.renamed(entry.text, library.copy_name(entry.name, taken)),
+                         note=f"A copy of “{entry.name}”. Nothing is saved until you press Save.")
+
+
+async def _checked(db, text: str, other_than: str | None = None):
+    """The template `text` describes and its problems: the checker's, and a
+    name another template already has."""
+    from .template_format import Problem, TemplateError, check
+    try:
+        template = check(text)
+    except TemplateError as exc:
+        return None, exc.problems
+    if (await library.name_taken(db, template.name, other_than)
+            or any(e.name == template.name for e in library.bundled())):
+        line = next((i for i, ln in enumerate(text.splitlines(), 1) if ln.startswith("name:")), 0)
+        return None, [Problem(line, f"A template is already named “{template.name}”; give this one its own name.")]
+    return template, []
+
+
+def _typed_text(form) -> str:
+    return (form.get("text") or "").replace("\r\n", "\n")
+
+
+@prometheus_bp.post("/templates/new")
+@require_role(UserRole.admin)
+async def create_template():
+    form = await request.form
+    text = _typed_text(form)
+    async with current_app.db_sessionmaker() as db:
+        async with db.begin():
+            template, problems = await _checked(db, text)
+            if problems:
+                return await _editor(text=text, problems=problems, status=422, saving=True)
+            row = PrometheusWidgetTemplate(name=template.name, text=text)
+            db.add(row)
+            await db.flush()
+            row_id = row.id
+    await log_audit(current_app, *_actor(), "prometheus.template_created", entity_type="prometheus_template",
+                    entity_id=row_id, detail={"name": template.name, "version": template.version})
+    return redirect(f"/plugins/prometheus/templates/custom:{row_id}")
+
+
+UPLOAD_LIMIT = 64 * 1024
+
+
+@prometheus_bp.post("/templates/upload")
+@require_role(UserRole.admin)
+async def upload_template():
+    """A .yaml file read into a new editor. Nothing is saved until Save."""
+    files = await request.files
+    upload = files.get("file")
+    if upload is None or not upload.filename:
+        return await _editor(problems=["Choose a .yaml file to upload."], status=422)
+    raw = upload.read(UPLOAD_LIMIT + 1)
+    if len(raw) > UPLOAD_LIMIT:
+        return await _editor(problems=[f"A template is at most {UPLOAD_LIMIT // 1024} KB."], status=422)
+    try:
+        text = raw.decode("utf-8-sig").replace("\r\n", "\n")
+    except UnicodeDecodeError:
+        return await _editor(problems=[f"{upload.filename} isn't a text file (UTF-8)."], status=422)
+    return await _editor(text=text, note=f"Read from {upload.filename}. Nothing is saved until you press Save.")
+
+
+@prometheus_bp.get("/templates/<key>")
+@require_role(UserRole.viewer)
+async def template_detail(key: str):
+    async with current_app.db_sessionmaker() as db:
+        entry = await _entry_or_404(db, key)
+    return await _editor(entry, problems=entry.problems)
+
+
+async def _custom_or_404(db, template_id: str) -> PrometheusWidgetTemplate:
+    row = await db.get(PrometheusWidgetTemplate, template_id)
+    if row is None:
+        abort(404)
+    return row
+
+
+@prometheus_bp.post("/templates/custom:<template_id>/edit")
+@require_role(UserRole.admin)
+async def save_template(template_id: str):
+    form = await request.form
+    text = _typed_text(form)
+    async with current_app.db_sessionmaker() as db:
+        async with db.begin():
+            row = await _custom_or_404(db, template_id)
+            template, problems = await _checked(db, text, other_than=row.id)
+            if problems:
+                entry = library.custom_entry(row)
+                return await _editor(entry, text=text, problems=problems, status=422, saving=True)
+            old_name = row.name
+            row.name, row.text = template.name, text
+            row.updated_at = datetime.now(timezone.utc)
+    detail = {"name": template.name, "version": template.version}
+    if old_name != template.name:
+        detail["old_name"] = old_name
+    await log_audit(current_app, *_actor(), "prometheus.template_updated", entity_type="prometheus_template",
+                    entity_id=template_id, detail=detail)
+    return redirect(f"/plugins/prometheus/templates/custom:{template_id}")
+
+
+@prometheus_bp.post("/templates/custom:<template_id>/delete")
+@require_role(UserRole.admin)
+async def delete_template(template_id: str):
+    async with current_app.db_sessionmaker() as db:
+        async with db.begin():
+            row = await _custom_or_404(db, template_id)
+            name = row.name
+            await db.delete(row)
+    await log_audit(current_app, *_actor(), "prometheus.template_deleted", entity_type="prometheus_template",
+                    entity_id=template_id, detail={"name": name})
+    return redirect("/plugins/prometheus/templates/")
+
+
+@prometheus_bp.get("/templates/<key>/export")
+@require_role(UserRole.viewer)
+async def export_template(key: str):
+    """The template's YAML as a file, to share or keep."""
+    async with current_app.db_sessionmaker() as db:
+        entry = await _entry_or_404(db, key)
+    return Response(entry.text, mimetype="application/yaml", headers={
+        "Content-Disposition": f'attachment; filename="{library.file_name(entry.name)}"'})
+
+
+@prometheus_bp.post("/templates/preview")
+@require_role(UserRole.viewer)
+async def preview_template():
+    """The editor's live preview: what's typed, checked, and drawn over a
+    target's stored values as the widget would draw it. Saves nothing."""
+    from .template_format import TemplateError, check
+    form = await request.form
+    text = _typed_text(form)
+    now = datetime.now(timezone.utc)
+    try:
+        template = check(text) if text.strip() else None
+    except TemplateError as exc:
+        return await render_template("prometheus/_preview.html", problems=[str(p) for p in exc.problems],
+                                     view=None, target=None)
+    if template is None:
+        return await render_template("prometheus/_preview.html", problems=[], view=None, target=None,
+                                     empty=True)
+    async with current_app.db_sessionmaker() as db:
+        target = await db.get(PrometheusTarget, form.get("target") or "") if form.get("target") else None
+        if target is None:
+            return await render_template("prometheus/_preview.html", problems=[], view=None, target=None,
+                                         template=template)
+        shown = await draw(db, template, target, now)
+    return await render_template("prometheus/_preview.html", problems=[], view=shown, template=template,
+                                 target=view_target(target, now), spark_hours=SPARK_HOURS)

plugins/prometheus/scheduler.py

@@ -0,0 +1,163 @@
+# plugins/prometheus/scheduler.py
+from __future__ import annotations
+
+import asyncio
+import logging
+import time
+from datetime import datetime, timezone
+from typing import TYPE_CHECKING
+
+from sqlalchemy import select
+
+from steward.core.counter_rates import CounterRates
+from steward.core.scheduler import ScheduledTask
+
+if TYPE_CHECKING:
+    from quart import Quart
+
+logger = logging.getLogger(__name__)
+
+# How often the scheduler looks for due targets. A target is scraped within
+# one tick of its interval, so this is the cadence's worst-case lateness.
+TICK_SECONDS = 5
+
+# A scrape past this is given up, whatever the client's own timeouts did:
+# a slow drip of bytes can keep each read inside its timeout for ever.
+SCRAPE_DEADLINE_SECONDS = 30
+
+# Counters' previous samples, for their rates.
+RATES = CounterRates()
+
+
+def make_scrape_task(app: "Quart") -> ScheduledTask:
+    async def tick() -> None:
+        await scrape_due(app)
+
+    return ScheduledTask(name="prometheus_scrape", coro_factory=tick,
+                         interval_seconds=TICK_SECONDS, run_on_startup=True)
+
+
+def is_due(target, now: datetime) -> bool:
+    """Whether a target's interval has elapsed since its last scrape, with
+    half a tick of slack so tick jitter doesn't push it back a whole tick."""
+    if not target.enabled:
+        return False
+    if target.last_scraped_at is None:
+        return True
+    elapsed = (now - target.last_scraped_at).total_seconds()
+    return elapsed >= target.interval_seconds - TICK_SECONDS / 2
+
+
+async def scrape_due(app: "Quart", scrape=None) -> None:
+    """Scrape every due target at once, then write each one's outcome to its
+    row and its series to the metric tables. `scrape` stands in for
+    scraper.scrape in tests."""
+    from steward.core.crypto import decrypt_secret
+
+    from .models import PrometheusTarget
+    from .scraper import ScrapeError
+    from .scraper import scrape as fetch
+
+    scrape = scrape or fetch
+    now = datetime.now(timezone.utc)
+    async with app.db_sessionmaker() as session:
+        targets = [t for t in (await session.execute(select(PrometheusTarget))).scalars().all()
+                   if is_due(t, now)]
+    if not targets:
+        return
+
+    async def one(t):
+        started = time.monotonic()
+        token = decrypt_secret(t.token, context=f"prometheus target {t.name} token")
+        try:
+            families = await asyncio.wait_for(scrape(t.url, token, t.verify_tls),
+                                              SCRAPE_DEADLINE_SECONDS)
+            return families, (time.monotonic() - started) * 1000
+        except TimeoutError:
+            return ScrapeError(f"Gave up after {SCRAPE_DEADLINE_SECONDS} s."), \
+                (time.monotonic() - started) * 1000
+        except Exception as exc:  # noqa: BLE001 — every failure is the target's outcome
+            return exc, (time.monotonic() - started) * 1000
+
+    results = await asyncio.gather(*(one(t) for t in targets))
+    await record(app, targets, results, datetime.now(timezone.utc))
+
+
+async def forget_gone(session, target: str, kept: set[tuple[str, str]]) -> None:
+    """Drop the newest value of each series of `target` that its last scrape
+    didn't keep: the endpoint stopped reporting it (an old build_info after an
+    upgrade, a deleted job), or a filter or the cap now leaves it out. Pages,
+    widgets and query_metrics read the newest values, so they show what the
+    endpoint says now. History stays for charts and ages out with retention.
+    Alert rules keep their own last-sample time, so one on a gone series
+    still goes no_data."""
+    from sqlalchemy import delete, tuple_
+
+    from steward.models.metrics import PluginMetricLatest
+
+    from .service import SOURCE
+
+    L = PluginMetricLatest
+    stored = (await session.execute(select(L.resource_name, L.metric_name).where(
+        L.source_module == SOURCE,
+        (L.resource_name == target) | L.resource_name.startswith(target + ":", autoescape=True)))).all()
+    gone = [(r, m) for r, m in stored if (r, m) not in kept]
+    if gone:
+        await session.execute(delete(L).where(
+            L.source_module == SOURCE, tuple_(L.resource_name, L.metric_name).in_(gone)))
+
+
+async def record(app: "Quart", targets, results, now: datetime) -> None:
+    """Each target's outcome on its row, and its kept series (counters as
+    rates) plus its own is_up / scrape_ms / series counts as metrics. After a
+    good scrape, the series it no longer has are forgotten (forget_gone); a
+    failed scrape says nothing about them, so they stay."""
+    from steward.core.alerts import MetricBatch
+
+    from .models import PrometheusTarget
+    from .scraper import ScrapeError
+    from .service import (IS_UP, OWN_METRICS, SCRAPE_MS, SERIES_CAPPED, SERIES_KEPT, SOURCE,
+                          map_families)
+
+    async with app.db_sessionmaker() as session:
+        async with session.begin():
+            batch = MetricBatch()
+            answered: list[tuple[str, set[tuple[str, str]]]] = []
+            for target, (result, ms) in zip(targets, results):
+                row = await session.get(PrometheusTarget, target.id)
+                if row is None:          # deleted while it was being scraped
+                    continue
+                row.last_scraped_at = now
+                if isinstance(result, BaseException):
+                    if not isinstance(result, ScrapeError):
+                        logger.warning("Prometheus scrape of %s failed: %r", target.name, result)
+                    row.last_error = str(result) or "The scrape failed."
+                    batch.add(source_module=SOURCE, resource_name=row.name, metric_name=IS_UP,
+                              value=0.0, recorded_at=now)
+                    continue
+                mapped = map_families(row.name, result, list(row.keep or []), list(row.drop or []),
+                                      row.series_cap)
+                row.last_error = None
+                row.last_ok_at = now
+                row.last_kept = len(mapped.series)
+                row.last_dropped = mapped.dropped
+                row.last_capped = mapped.capped
+                # A counter's first scrape gives no rate yet; it is still kept.
+                answered.append((row.name, {(x.resource, x.metric) for x in mapped.series}
+                                 | {(row.name, m) for m in OWN_METRICS}))
+                for series in mapped.series:
+                    value = series.value
+                    if series.counter:
+                        value = RATES.rate((row.id, series.resource, series.metric), value, now)
+                        if value is None:
+                            continue
+                    batch.add(source_module=SOURCE, resource_name=series.resource,
+                              metric_name=series.metric, value=value, recorded_at=now)
+                for metric_name, value in [(IS_UP, 1.0), (SCRAPE_MS, ms),
+                                           (SERIES_KEPT, len(mapped.series)),
+                                           (SERIES_CAPPED, mapped.capped)]:
+                    batch.add(source_module=SOURCE, resource_name=row.name,
+                              metric_name=metric_name, value=value, recorded_at=now)
+            await batch.record(session)
+            for name, kept in answered:
+                await forget_gone(session, name, kept)

plugins/prometheus/scraper.py

@@ -0,0 +1,58 @@
+# plugins/prometheus/scraper.py
+"""Fetching one /metrics endpoint. Every failure comes back as one sentence
+a page can show, never a stack trace."""
+from __future__ import annotations
+
+import httpx
+
+from steward.core import prometheus
+
+TIMEOUT_SECONDS = 10.0
+# A /metrics page past this is not one to store a series of; reading stops.
+MAX_BYTES = 20 * 1024 * 1024
+
+ACCEPT = "text/plain;version=0.0.4;q=1,*/*;q=0.1"
+
+
+class ScrapeError(Exception):
+    """Why a scrape got nothing, said to the person."""
+
+
+async def scrape(url: str, token: str = "", verify_tls: bool = True,
+                 transport: httpx.AsyncBaseTransport | None = None) -> dict[str, prometheus.Family]:
+    """The endpoint's families. `transport` stands in for the network in tests."""
+    headers = {"Accept": ACCEPT}
+    if token:
+        headers["Authorization"] = f"Bearer {token}"
+    try:
+        async with httpx.AsyncClient(timeout=TIMEOUT_SECONDS, verify=verify_tls,
+                                     transport=transport, follow_redirects=False) as client:
+            async with client.stream("GET", url, headers=headers) as response:
+                if response.status_code in (401, 403):
+                    raise ScrapeError(f"Answered {response.status_code}: the token was refused. "
+                                      "Check the token, or make a new one on the service.")
+                if response.status_code != 200:
+                    raise ScrapeError(f"Answered {response.status_code} {response.reason_phrase}, not metrics.")
+                body = bytearray()
+                async for chunk in response.aiter_bytes():
+                    body.extend(chunk)
+                    if len(body) > MAX_BYTES:
+                        raise ScrapeError(f"The page is over {MAX_BYTES // (1024 * 1024)} MB; "
+                                          "is this a /metrics address?")
+    except ScrapeError:
+        raise
+    except httpx.TimeoutException:
+        raise ScrapeError(f"No answer within {TIMEOUT_SECONDS:.0f} s.") from None
+    except httpx.ConnectError as exc:
+        raise ScrapeError(f"Couldn't connect: {_reason(exc)}.") from None
+    except httpx.HTTPError as exc:
+        raise ScrapeError(f"The request failed: {_reason(exc)}.") from None
+    families = prometheus.parse(body.decode("utf-8", errors="replace"))
+    if not families:
+        raise ScrapeError("It answered, but with no metrics in Prometheus format.")
+    return families
+
+
+def _reason(exc: Exception) -> str:
+    text = str(exc).strip() or type(exc).__name__
+    return text.rstrip(".")

plugins/prometheus/service.py

@@ -0,0 +1,269 @@
+"""Prometheus targets: the form a person edits, the row it becomes, how a
+scrape's series become Steward metrics, and what a page says about the
+newest scrape."""
+from __future__ import annotations
+
+import math
+from dataclasses import dataclass, field
+from datetime import datetime, timezone
+from fnmatch import fnmatchcase
+
+from steward.core import prometheus
+from steward.core.crypto import encrypt_secret
+from steward.core.formatting import metric_value
+
+from .models import PrometheusTarget
+
+SOURCE = "prometheus"
+
+# What each target records about its own scrape, under its own name.
+IS_UP = "is_up"
+SCRAPE_MS = "scrape_ms"
+SERIES_KEPT = "series_kept"
+SERIES_CAPPED = "series_capped"
+OWN_METRICS = (IS_UP, SCRAPE_MS, SERIES_KEPT, SERIES_CAPPED)
+
+# Column widths of plugin_metrics; a longer name can't be stored.
+RESOURCE_MAX = 255
+METRIC_MAX = 128
+
+INTERVAL_MIN, INTERVAL_MAX = 10, 3600
+CAP_MIN, CAP_MAX = 1, 10000
+
+
+class FormError(ValueError):
+    """What is wrong with a submitted target form, said to the person."""
+
+
+# ── Series → metrics ─────────────────────────────────────────────────────────
+
+def wanted(name: str, keep: list[str], drop: list[str]) -> bool:
+    """Whether a family is stored: it matches a keep pattern (or there are
+    none) and no drop pattern. Drop wins over keep."""
+    if keep and not any(fnmatchcase(name, p) for p in keep):
+        return False
+    return not any(fnmatchcase(name, p) for p in drop)
+
+
+def resource_for(target: str, labels: tuple[tuple[str, str], ...]) -> str:
+    """The resource a series is stored under: the target's name, then its
+    labels as key=value, so "Archives:target=Steward,state=late". A series
+    without labels is the target itself."""
+    if not labels:
+        return target
+    return target + ":" + ",".join(f"{k}={v}" for k, v in labels)
+
+
+def rate_name(sample_name: str) -> str:
+    """A counter is stored as a per-second rate: requests_total →
+    requests_per_sec, job_seconds_sum → job_seconds_sum_per_sec."""
+    base = sample_name[: -len("_total")] if sample_name.endswith("_total") else sample_name
+    return base + "_per_sec"
+
+
+def label_pairs(labels: str) -> list[tuple[str, str]]:
+    """resource_for's labels read back: "target=S,state=late" →
+    [("target", "S"), ("state", "late")]. A piece without "=" was part of the
+    value before it, a value that held a comma."""
+    pairs: list[tuple[str, str]] = []
+    for piece in labels.split(",") if labels else []:
+        key, eq, value = piece.partition("=")
+        if eq and key:
+            pairs.append((key, value))
+        elif pairs:
+            pairs[-1] = (pairs[-1][0], pairs[-1][1] + "," + piece)
+    return pairs
+
+
+def state_label(metric: str, pairs: list[tuple[str, str]]) -> str | None:
+    """Which label names a state set's state: `state`, or the metric's own
+    name as OpenMetrics writes a stateset (foo{foo="a"} 1)."""
+    keys = {k for k, _ in pairs}
+    for key in ("state", metric):
+        if key in keys:
+            return key
+    return None
+
+
+def fold_states(metric: str, rows: list[dict]) -> list[dict] | None:
+    """A state set's series folded to one row per thing it describes.
+
+    A state set is one series per possible state, the current one 1 and the
+    rest 0: Archives' fabledarchives_backup_target_state{target, state}. Each
+    row in `rows` has `labels` (resource_for's form), `value` and `resource`.
+    A metric is folded when its name ends in _state (or its state label is
+    its own name) and every series carries the state label and reads 0 or 1;
+    anything else is None, and shown series by series. The name matters: a
+    count split by a `state` label (fabledarchives_jobs{state="waiting"})
+    often reads 0 or 1 too, and is not a state set.
+
+    Each folded row: `labels` without the state, `state` (the one set to 1;
+    several are joined, none is ""), `states` (every state, in order) and
+    `resource` (the current state's series, else the first)."""
+    if not rows:
+        return None
+    sets: dict[str, list[tuple[str, dict]]] = {}
+    for row in rows:
+        pairs = label_pairs(row["labels"])
+        key = state_label(metric, pairs)
+        if key is None or row["value"] not in (0, 1):
+            return None
+        if key == "state" and not metric.endswith("_state"):
+            return None
+        rest = ",".join(f"{k}={v}" for k, v in pairs if k != key)
+        sets.setdefault(rest, []).append((dict(pairs)[key], row))
+    folded = []
+    for rest, members in sorted(sets.items()):
+        members.sort(key=lambda m: m[0])
+        current = [(s, r) for s, r in members if r["value"] == 1]
+        folded.append({"labels": rest, "state": ", ".join(s for s, _ in current),
+                       "states": [s for s, _ in members],
+                       "resource": (current or members)[0][1]["resource"]})
+    return folded
+
+
+def shown_value(metric: str, value: float | None) -> dict:
+    """A series' value in its unit, from the Prometheus naming conventions.
+    A *_timestamp_seconds is a moment: it comes back as `moment` for the page
+    to show as "3m ago" in the reader's zone; 0 there means "never"."""
+    if value is None:
+        return {"text": "—", "moment": None}
+    if metric.endswith("_timestamp_seconds"):
+        if value <= 0:
+            return {"text": "never", "moment": None}
+        return {"text": "", "moment": datetime.fromtimestamp(value, timezone.utc)}
+    return {"text": metric_value(metric, value), "moment": None}
+
+
+@dataclass
+class Series:
+    resource: str
+    metric: str
+    value: float
+    counter: bool
+
+
+@dataclass
+class Mapped:
+    series: list[Series] = field(default_factory=list)
+    dropped: int = 0     # left out by the keep/drop patterns
+    capped: int = 0      # past the series cap, or a name too long to store
+
+
+def map_families(target: str, families: dict[str, prometheus.Family], keep: list[str],
+                 drop: list[str], cap: int) -> Mapped:
+    """The series of one scrape that Steward stores, in the endpoint's order.
+
+    Gauges and untyped samples are stored as they are; counters (and a
+    histogram's or summary's _sum and _count) are marked to become rates.
+    Histogram buckets and summary quantiles are skipped: Steward stores one
+    number per series, not a distribution. NaN and infinite values are
+    skipped as unmeasured."""
+    out = Mapped()
+    for family in families.values():
+        if not wanted(family.name, keep, drop):
+            out.dropped += len(family.samples)
+            continue
+        for sample in family.samples:
+            counter = family.type == "counter"
+            if family.type in ("histogram", "summary"):
+                if not sample.name.endswith(("_sum", "_count")):
+                    continue
+                counter = True
+            if math.isnan(sample.value) or math.isinf(sample.value):
+                continue
+            metric = rate_name(sample.name) if counter else sample.name
+            resource = resource_for(target, sample.labels)
+            if len(out.series) >= cap or len(metric) > METRIC_MAX or len(resource) > RESOURCE_MAX:
+                out.capped += 1
+                continue
+            out.series.append(Series(resource, metric, sample.value, counter))
+    return out
+
+
+# ── The form ─────────────────────────────────────────────────────────────────
+
+def patterns(text: str) -> list[str]:
+    """Name patterns from a textarea: one per line or comma-separated."""
+    out: list[str] = []
+    for line in (text or "").replace(",", "\n").splitlines():
+        p = line.strip()
+        if p and p not in out:
+            out.append(p)
+    return out
+
+
+def form_values(target: PrometheusTarget | None) -> dict:
+    """The form's fields for an existing target. The token is never sent back
+    to the browser; a blank token field on save keeps the stored one."""
+    if target is None:
+        return {"interval_seconds": 60, "series_cap": 1000, "verify_tls": True, "enabled": True,
+                "keep": "", "drop": ""}
+    return {"name": target.name, "url": target.url, "interval_seconds": target.interval_seconds,
+            "series_cap": target.series_cap, "verify_tls": target.verify_tls,
+            "enabled": target.enabled, "keep": "\n".join(target.keep or []),
+            "drop": "\n".join(target.drop or []), "has_token": bool(target.token)}
+
+
+def _bounded(form, key: str, default: int, low: int, high: int, says: str) -> int:
+    try:
+        value = int(form.get(key) or default)
+    except ValueError:
+        raise FormError(says) from None
+    if not low <= value <= high:
+        raise FormError(says)
+    return value
+
+
+def parse_form(form, existing: PrometheusTarget | None = None) -> dict:
+    """The submitted form as the columns to store, the token encrypted.
+    Raises FormError saying what to fix."""
+    name = (form.get("name") or "").strip()
+    url = (form.get("url") or "").strip()
+    if not name:
+        raise FormError("Give the target a name.")
+    if len(name) > 100 or any(c in name for c in "/:,"):
+        raise FormError("A name is up to 100 characters, without / : or a comma.")
+    if not url.startswith(("http://", "https://")) or len(url) > 500:
+        raise FormError("The address starts with http:// or https://, e.g. http://archives:8081/metrics.")
+    interval = _bounded(form, "interval_seconds", 60, INTERVAL_MIN, INTERVAL_MAX,
+                        f"Scrape every {INTERVAL_MIN} to {INTERVAL_MAX} seconds.")
+    cap = _bounded(form, "series_cap", 1000, CAP_MIN, CAP_MAX,
+                   f"The series cap is between {CAP_MIN} and {CAP_MAX}.")
+    remove_token = bool(form.get("clear_token"))
+    typed = (form.get("token") or "").strip()
+    if typed:
+        token = encrypt_secret(typed)
+    elif existing is not None and not remove_token:
+        token = existing.token
+    else:
+        token = ""
+    return {"name": name, "url": url, "token": token,
+            "verify_tls": bool(form.get("verify_tls")), "interval_seconds": interval,
+            "keep": patterns(form.get("keep") or ""), "drop": patterns(form.get("drop") or ""),
+            "series_cap": cap, "enabled": bool(form.get("enabled"))}
+
+
+# ── What the newest scrape says ──────────────────────────────────────────────
+
+# A target not scraped for this many intervals has stopped being scraped
+# (plugin turned off, scheduler stuck), whatever its last outcome said.
+STALE_SCRAPES = 3
+
+
+@dataclass
+class TargetState:
+    reach: str                 # up | down | stale | pending | off
+    says: str                  # one line, in words
+
+
+def target_state(target: PrometheusTarget, now: datetime) -> TargetState:
+    if not target.enabled:
+        return TargetState("off", "Not scraped: turned off.")
+    if target.last_scraped_at is None:
+        return TargetState("pending", "Not scraped yet.")
+    if (now - target.last_scraped_at).total_seconds() > STALE_SCRAPES * max(target.interval_seconds, 1):
+        return TargetState("stale", "Not scraped recently: the Prometheus plugin may be turned off.")
+    if target.last_error:
+        return TargetState("down", target.last_error)
+    return TargetState("up", "Answering.")

plugins/prometheus/template_format.py

@@ -0,0 +1,458 @@
+"""Prometheus widget templates: the YAML a person writes, read and checked.
+
+A template says which of a target's series a dashboard widget shows and
+how. It is data only: it names series and lays them out in blocks, and
+Steward draws them (template_view.py). It can't add a visual, a colour or a
+formula; a new block is a Steward release (System 181's ruling). The
+format, for people writing one: docs/plugins/prometheus-widgets.md.
+
+    name: Fabled Archives
+    version: 1
+    expects: fabledarchives_
+    blocks:
+      - states:
+          title: Backup targets
+          metric: fabledarchives_backup_target_state
+          name: target
+          healthy: [fresh]
+          quiet: [disabled, waiting]
+          warning: [late]
+      - values:
+          items:
+            - {label: Disk free, metric: fabledarchives_data_disk_free_bytes,
+               of: fabledarchives_data_disk_size_bytes, warn_below: 10%}
+
+check(text) returns the Template, or every problem found, each with its
+line and a sentence saying what to change.
+"""
+from __future__ import annotations
+
+import re
+from dataclasses import dataclass, field
+
+import yaml
+
+FORMAT = 1
+MAX_BYTES = 64 * 1024
+MAX_BLOCKS = 20
+MAX_ITEMS = 30
+
+_METRIC = re.compile(r"^[a-zA-Z_:][a-zA-Z0-9_:]*$")
+_LABEL = re.compile(r"^[a-zA-Z_][a-zA-Z0-9_]*$")
+_DURATION = re.compile(r"^(\d+(?:\.\d+)?)\s*([smhd]?)$")
+_SECONDS = {"": 1, "s": 1, "m": 60, "h": 3600, "d": 86400}
+
+# What a unit shows a value as; the name is the suffix metric_value reads.
+UNITS = {"bytes": "_bytes", "seconds": "_secs", "percent": "_pct", "per_sec": "_per_sec",
+         "count": "", "timestamp": "_timestamp_seconds"}
+
+
+@dataclass
+class Problem:
+    line: int
+    text: str
+
+    def __str__(self) -> str:
+        return f"Line {self.line}: {self.text}" if self.line else self.text
+
+
+@dataclass
+class Block:
+    kind: str        # states | values | times | table | info
+    title: str
+    line: int
+    spec: dict = field(default_factory=dict)
+
+
+@dataclass
+class Template:
+    name: str
+    version: int
+    description: str = ""
+    expects: str = ""
+    blocks: list[Block] = field(default_factory=list)
+    text: str = ""
+
+
+class TemplateError(ValueError):
+    def __init__(self, problems: list[Problem]):
+        self.problems = problems
+        super().__init__("; ".join(str(p) for p in problems))
+
+
+# ── YAML with line numbers ───────────────────────────────────────────────────
+
+class _Map(dict):
+    line = 0
+    lines: dict
+
+
+class _List(list):
+    line = 0
+    lines: list
+
+
+class _Loader(yaml.SafeLoader):
+    pass
+
+
+def _construct_map(loader, node):
+    loader.flatten_mapping(node)
+    out = _Map()
+    out.line, out.lines = node.start_mark.line + 1, {}
+    for key_node, value_node in node.value:
+        key = loader.construct_object(key_node, deep=True)
+        if not isinstance(key, str):
+            raise yaml.constructor.ConstructorError(
+                None, None, "a key here should be a word, like `metric`", key_node.start_mark)
+        if key in out:
+            raise yaml.constructor.ConstructorError(
+                None, None, f"`{key}` is given twice", key_node.start_mark)
+        out[key] = loader.construct_object(value_node, deep=True)
+        out.lines[key] = key_node.start_mark.line + 1
+    return out
+
+
+def _construct_list(loader, node):
+    out = _List(loader.construct_object(n, deep=True) for n in node.value)
+    out.line, out.lines = node.start_mark.line + 1, [n.start_mark.line + 1 for n in node.value]
+    return out
+
+
+_Loader.add_constructor(yaml.resolver.BaseResolver.DEFAULT_MAPPING_TAG, _construct_map)
+_Loader.add_constructor(yaml.resolver.BaseResolver.DEFAULT_SEQUENCE_TAG, _construct_list)
+
+
+def _load(text: str) -> tuple[object, list[Problem]]:
+    if len(text.encode()) > MAX_BYTES:
+        return None, [Problem(0, f"A template is at most {MAX_BYTES // 1024} KB.")]
+    try:
+        # An alias can make a small file expand without end; a template has no use for one.
+        for event in yaml.parse(text, Loader=yaml.SafeLoader):
+            if isinstance(event, yaml.AliasEvent):
+                return None, [Problem(event.start_mark.line + 1,
+                                      "Anchors and aliases (&name, *name) aren't allowed; write the value out.")]
+        return yaml.load(text, Loader=_Loader), []
+    except yaml.constructor.ConstructorError as exc:      # ours: a key twice, or not a word
+        mark = exc.problem_mark
+        return None, [Problem(mark.line + 1 if mark else 0, exc.problem[:1].upper() + exc.problem[1:] + ".")]
+    except yaml.MarkedYAMLError as exc:
+        mark = exc.problem_mark or exc.context_mark
+        return None, [Problem(mark.line + 1 if mark else 0, f"This isn't valid YAML: {exc.problem}.")]
+    except yaml.YAMLError as exc:
+        return None, [Problem(0, f"This isn't valid YAML: {exc}.")]
+
+
+# ── Values ───────────────────────────────────────────────────────────────────
+
+def duration(value) -> float | None:
+    """Seconds from 90, "90s", "15m", "26h" or "2d"; None if it isn't one."""
+    if isinstance(value, bool):
+        return None
+    if isinstance(value, (int, float)):
+        return float(value) if value >= 0 else None
+    m = _DURATION.match(str(value).strip().lower())
+    return float(m.group(1)) * _SECONDS[m.group(2)] if m else None
+
+
+@dataclass
+class Limit:
+    """A threshold: a number, or a share of the item's `of` ("10%")."""
+    value: float
+    share: bool = False
+
+
+def limit(value) -> Limit | None:
+    if isinstance(value, bool):
+        return None
+    if isinstance(value, (int, float)):
+        return Limit(float(value))
+    text = str(value).strip()
+    if text.endswith("%"):
+        try:
+            return Limit(float(text[:-1]), share=True)
+        except ValueError:
+            return None
+    return None
+
+
+# ── The checker ──────────────────────────────────────────────────────────────
+
+class _Check:
+    def __init__(self):
+        self.problems: list[Problem] = []
+
+    def add(self, line: int, text: str) -> None:
+        self.problems.append(Problem(line, text))
+
+    def keys(self, m: _Map, allowed: tuple, required: tuple, where: str) -> bool:
+        ok = True
+        for key in m:
+            if key not in allowed:
+                self.add(m.lines[key], f"{where} has no `{key}`. It takes: {', '.join(allowed)}.")
+                ok = False
+        for key in required:
+            if key not in m:
+                self.add(m.line, f"{where} needs `{key}`.")
+                ok = False
+        return ok
+
+    def text(self, m: _Map, key: str, where: str, default: str = "") -> str:
+        if key not in m:
+            return default
+        value = m[key]
+        if isinstance(value, (dict, list)) or value is None:
+            self.add(m.lines[key], f"`{key}` in {where} should be text.")
+            return default
+        return str(value)
+
+    def metric(self, m: _Map, key: str, where: str) -> str:
+        value = m.get(key)
+        if not isinstance(value, str) or not _METRIC.match(value):
+            self.add(m.lines.get(key, m.line),
+                     f"`{key}` in {where} should be a metric name as Steward stores it, like "
+                     "fabledarchives_files (a counter is stored as <name>_per_sec).")
+            return ""
+        return value
+
+    def label(self, m: _Map, key: str, where: str) -> str:
+        if key not in m:
+            return ""
+        value = m[key]
+        if not isinstance(value, str) or not _LABEL.match(value):
+            self.add(m.lines[key], f"`{key}` in {where} should be a label name, like target.")
+            return ""
+        return value
+
+    def words(self, m: _Map, key: str, where: str) -> list[str]:
+        if key not in m:
+            return []
+        value = m[key]
+        if not isinstance(value, list) or not all(isinstance(v, (str, int)) for v in value):
+            self.add(m.lines[key], f"`{key}` in {where} should be a list, like [fresh, waiting].")
+            return []
+        if any(isinstance(v, bool) for v in value):
+            self.add(m.lines[key], f"`{key}` in {where} has a word YAML reads as true or false "
+                                   "(on, off, yes, no); put it in quotes, like 'off'.")
+            return []
+        return [str(v) for v in value]
+
+    def where(self, m: _Map, ctx: str) -> dict[str, str]:
+        if "where" not in m:
+            return {}
+        value = m["where"]
+        if not isinstance(value, dict) or not all(
+                isinstance(k, str) and _LABEL.match(k) and isinstance(v, (str, int, float))
+                and not isinstance(v, bool) for k, v in value.items()):
+            self.add(m.lines["where"], f"`where` in {ctx} should map labels to values, like "
+                                       "{kind: files}. Quote a value YAML would read as true or false.")
+            return {}
+        return {k: str(v) for k, v in value.items()}
+
+    def unit(self, m: _Map, ctx: str) -> str:
+        if "unit" not in m:
+            return ""
+        value = m["unit"]
+        if value not in UNITS:
+            self.add(m.lines["unit"], f"`unit` in {ctx} is one of: {', '.join(UNITS)}.")
+            return ""
+        return value
+
+    def span(self, m: _Map, key: str, ctx: str) -> float | None:
+        if key not in m:
+            return None
+        value = duration(m[key])
+        if value is None:
+            self.add(m.lines[key], f"`{key}` in {ctx} should be a span, like 26h, 90m or 2d.")
+        return value
+
+    def items(self, m: _Map, ctx: str) -> list[_Map]:
+        value = m.get("items")
+        if not isinstance(value, list) or not value:
+            self.add(m.lines.get("items", m.line), f"`items` in {ctx} should be a list of one or more items.")
+            return []
+        if len(value) > MAX_ITEMS:
+            self.add(m.lines["items"], f"{ctx} has {len(value)} items; at most {MAX_ITEMS}.")
+        out = []
+        for i, item in enumerate(value):
+            if not isinstance(item, dict):
+                self.add(value.lines[i], f"Each item in {ctx} should be a set of keys, like "
+                                         "{label: Disk free, metric: …}.")
+            else:
+                out.append(item)
+        return out
+
+
+# Each block: the keys it takes, those it needs, and its title when none is given.
+_BLOCKS = {
+    "states": (("title", "metric", "where", "name", "healthy", "quiet", "warning", "warn_past",
+                "crit_past", "columns"),
+               ("metric", "healthy"), "States"),
+    "values": (("title", "items"), ("items",), "Figures"),
+    "times": (("title", "items"), ("items",), "Times"),
+    "table": (("title", "metric", "where", "rows", "columns", "unit"), ("metric", "rows", "columns"),
+              "Table"),
+    "info": (("title", "metric", "where", "labels"), ("metric",), "About"),
+}
+_VALUE_ITEM = ("label", "metric", "where", "name", "of", "unit", "warn_above", "crit_above",
+               "warn_below", "crit_below", "sparkline")
+_TIME_ITEM = ("label", "metric", "where", "name", "warn_after", "crit_after")
+_COLUMN = ("label", "metric", "unit")
+
+
+def _states(c: _Check, m: _Map, ctx: str) -> dict:
+    spec = {"metric": c.metric(m, "metric", ctx), "where": c.where(m, ctx), "name": c.label(m, "name", ctx),
+            "healthy": c.words(m, "healthy", ctx), "quiet": c.words(m, "quiet", ctx),
+            "warning": c.words(m, "warning", ctx), "columns": [],
+            # A deadline per row (a *_timestamp_seconds): once it has passed, a
+            # healthy row reads as a warning or a problem. A job that says
+            # "running" for days is still overdue.
+            "warn_past": c.metric(m, "warn_past", ctx) if "warn_past" in m else "",
+            "crit_past": c.metric(m, "crit_past", ctx) if "crit_past" in m else ""}
+    if "healthy" in m and not spec["healthy"] and isinstance(m["healthy"], list):
+        c.add(m.lines["healthy"], f"`healthy` in {ctx} needs at least one state.")
+    seen: dict[str, str] = {}
+    for key in ("healthy", "quiet", "warning"):
+        for state in spec[key]:
+            if state in seen:
+                c.add(m.lines[key], f"“{state}” is in both `{seen[state]}` and `{key}` in {ctx}.")
+            seen[state] = key
+    if "columns" in m:
+        cols = m["columns"]
+        if not isinstance(cols, list):
+            c.add(m.lines["columns"], f"`columns` in {ctx} should be a list of {{label, metric}}.")
+        else:
+            for i, col in enumerate(cols):
+                where = f"a column of {ctx}"
+                if not isinstance(col, dict):
+                    c.add(cols.lines[i], f"Each of `columns` in {ctx} should be {{label, metric}}.")
+                    continue
+                if c.keys(col, _COLUMN, ("label", "metric"), where):
+                    spec["columns"].append({"label": c.text(col, "label", where),
+                                            "metric": c.metric(col, "metric", where),
+                                            "unit": c.unit(col, where)})
+    return spec
+
+
+def _value_item(c: _Check, item: _Map, ctx: str) -> dict | None:
+    where = f"an item of {ctx}"
+    if not c.keys(item, _VALUE_ITEM, ("label", "metric"), where):
+        return None
+    spec = {"label": c.text(item, "label", where), "metric": c.metric(item, "metric", where),
+            "where": c.where(item, where), "name": c.label(item, "name", where),
+            "of": c.metric(item, "of", where) if "of" in item else "", "unit": c.unit(item, where),
+            "sparkline": False}
+    if "sparkline" in item:
+        if not isinstance(item["sparkline"], bool):
+            c.add(item.lines["sparkline"], f"`sparkline` in {where} is true or false.")
+        spec["sparkline"] = item["sparkline"] is True
+    for key in ("warn_above", "crit_above", "warn_below", "crit_below"):
+        spec[key] = None
+        if key in item:
+            lim = limit(item[key])
+            if lim is None:
+                c.add(item.lines[key], f"`{key}` in {where} should be a number, or a share like 10%.")
+            elif lim.share and not spec["of"]:
+                c.add(item.lines[key], f"`{key}` in {where} is a share ({item[key]}), which needs `of`: "
+                                       "the metric it is a share of.")
+            spec[key] = lim
+    return spec
+
+
+def _time_item(c: _Check, item: _Map, ctx: str) -> dict | None:
+    where = f"an item of {ctx}"
+    if not c.keys(item, _TIME_ITEM, ("label", "metric"), where):
+        return None
+    spec = {"label": c.text(item, "label", where), "metric": c.metric(item, "metric", where),
+            "where": c.where(item, where), "name": c.label(item, "name", where),
+            "warn_after": c.span(item, "warn_after", where), "crit_after": c.span(item, "crit_after", where)}
+    if spec["warn_after"] and spec["crit_after"] and spec["crit_after"] < spec["warn_after"]:
+        c.add(item.lines["crit_after"], f"`crit_after` in {where} should be later than `warn_after`.")
+    return spec
+
+
+def _block(c: _Check, raw, line: int, n: int) -> Block | None:
+    if not isinstance(raw, dict) or len(raw) != 1:
+        c.add(line, f"Block {n} should be one kind with its settings under it, like "
+                    f"`- states: {{metric: …}}`. The kinds are: {', '.join(_BLOCKS)}.")
+        return None
+    kind, body = next(iter(raw.items()))
+    if kind not in _BLOCKS:
+        c.add(line, f"Block {n} is “{kind}”; the kinds are: {', '.join(_BLOCKS)}.")
+        return None
+    if not isinstance(body, dict):
+        c.add(line, f"Block {n} ({kind}) should have its settings under it, like `metric: …`.")
+        return None
+    allowed, required, default_title = _BLOCKS[kind]
+    ctx = f"block {n} ({kind})"
+    if not c.keys(body, allowed, required, ctx.capitalize()):
+        return None
+    title = c.text(body, "title", ctx, default_title)
+    if kind == "states":
+        spec = _states(c, body, ctx)
+    elif kind in ("values", "times"):
+        reader = _value_item if kind == "values" else _time_item
+        spec = {"items": [s for s in (reader(c, i, ctx) for i in c.items(body, ctx)) if s]}
+    elif kind == "table":
+        spec = {"metric": c.metric(body, "metric", ctx), "where": c.where(body, ctx),
+                "rows": c.label(body, "rows", ctx), "columns": c.label(body, "columns", ctx),
+                "unit": c.unit(body, ctx)}
+    else:
+        spec = {"metric": c.metric(body, "metric", ctx), "where": c.where(body, ctx),
+                "labels": c.words(body, "labels", ctx)}
+    return Block(kind, title, line, spec)
+
+
+_TOP = ("name", "description", "version", "format", "expects", "blocks")
+
+
+def check(text: str) -> Template:
+    """The template `text` describes. Raises TemplateError with every problem
+    found, each with its line, in the order they appear."""
+    raw, problems = _load(text)
+    if problems:
+        raise TemplateError(problems)
+    c = _Check()
+    if not isinstance(raw, dict):
+        raise TemplateError([Problem(1, "A template starts with its keys: name, version and blocks.")])
+    c.keys(raw, _TOP, ("name", "version", "blocks"), "A template")
+    name = c.text(raw, "name", "the template").strip()
+    if "name" in raw and not name:
+        c.add(raw.lines["name"], "Give the template a name.")
+    if len(name) > 80:
+        c.add(raw.lines["name"], "A name is at most 80 characters.")
+    version = raw.get("version")
+    if "version" in raw and (isinstance(version, bool) or not isinstance(version, int) or version < 1):
+        c.add(raw.lines["version"], "`version` is the template's own version: a whole number from 1. "
+                                    "Raise it when you change a template you have shared.")
+    if "format" in raw and raw["format"] != FORMAT:
+        c.add(raw.lines["format"], f"This Steward reads template format {FORMAT}.")
+    expects = c.text(raw, "expects", "the template")
+    if expects and not re.match(r"^[a-zA-Z_:][a-zA-Z0-9_:]*$", expects):
+        c.add(raw.lines["expects"], "`expects` should be the start of the metric names it reads, "
+                                    "like fabledarchives_.")
+    blocks: list[Block] = []
+    raw_blocks = raw.get("blocks")
+    if "blocks" in raw:
+        if not isinstance(raw_blocks, list) or not raw_blocks:
+            c.add(raw.lines["blocks"], "`blocks` should be a list of one or more blocks.")
+        else:
+            if len(raw_blocks) > MAX_BLOCKS:
+                c.add(raw.lines["blocks"], f"A template has at most {MAX_BLOCKS} blocks.")
+            for i, b in enumerate(raw_blocks):
+                block = _block(c, b, raw_blocks.lines[i], i + 1)
+                if block:
+                    blocks.append(block)
+    if c.problems:
+        raise TemplateError(sorted(c.problems, key=lambda p: p.line))
+    return Template(name=name, version=version, description=c.text(raw, "description", "the template"),
+                    expects=expects, blocks=blocks, text=text)
+
+
+def problems(text: str) -> list[Problem]:
+    """Every problem with `text`; empty when it is a template."""
+    try:
+        check(text)
+    except TemplateError as exc:
+        return exc.problems
+    return []