refactor(docker): container CPU and memory live in the one metric store (#5752)

#5752 "Container CPU and memory history lives in the one metric store;
docker_metrics and docker_metrics_hourly are retired". Container cpu_pct and
mem_pct already went to plugin_metrics for alerting; now mem_usage_bytes does
too, and every snapshot in an ingest (backlog included) is recorded at its
own time. The container history chart, Busiest Containers, a service's
history and the MCP container tools read plugin_metrics under resource
"<host>/<container>", rolled up and pruned by the shared metric retention.

docker_012_drop_metrics drops both tables. Their history is not copied
(lesson #5221 "A data migration's cost is set by the live table"); the
store already holds 7 days of container CPU and memory %. Docker retention
now prunes only lifecycle events, and Settings -> Thresholds has one
"Metric retention" card for every source.

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

steward/backup/contents.py

@@ -40,7 +40,7 @@ EXCLUDED: frozenset[str] = frozenset({
     # Run history; its logs live on disk, not in the database.
     "ansible_runs",
     "docker_containers", "docker_disk_usage", "docker_events", "docker_host_status",
-    "docker_images", "docker_logs", "docker_metrics", "docker_metrics_hourly",
+    "docker_images", "docker_logs",
     "docker_swarm_nodes", "docker_swarm_services",
     "traefik_access_summaries", "traefik_certs", "traefik_global_stats", "traefik_metrics",
     "traefik_status_codes",

steward/core/cleanup.py

@@ -23,8 +23,9 @@ logger = logging.getLogger(__name__)
 
 # Time-series columns pruned at data.retention_days (#5155). A plugin adds its
 # own with prune_by_age() in setup(), so the core never imports plugin models.
-# Tables kept another way are not listed: plugin_metrics and the docker tables
-# are rolled up (below), and the audit log has its own, longer window.
+# Tables kept another way are not listed: plugin_metrics is rolled up (below),
+# the docker events and logs have their own windows, and the audit log has its
+# own, longer window.
 _AGED: list = [MonitorResult.checked_at, AnsibleRun.started_at, AlertEvent.transitioned_at,
                EmailOutbox.created_at]
 
@@ -40,9 +41,7 @@ def prune_by_age(column) -> None:
 _METRIC_COUNTS = [("raw_rows_rolled", "plugin_metrics: raw samples rolled into hourly"),
                   ("rollup_pruned", "plugin_metrics_hourly: past the rollup window"),
                   ("latest_pruned", "plugin_metric_latest: stopped reporting")]
-_DOCKER_COUNTS = [("raw_rows_rolled", "docker_metrics: raw samples rolled into hourly"),
-                  ("rollup_pruned", "docker_metrics_hourly: past the rollup window"),
-                  ("events_pruned", "docker_events: past the events window"),
+_DOCKER_COUNTS = [("events_pruned", "docker_events: past the events window"),
                   ("logs_age_pruned", "docker_logs: past the age window"),
                   ("logs_size_pruned", "docker_logs: over a container's size cap")]
 
@@ -142,13 +141,13 @@ async def _run_metrics_retention(session, now: datetime) -> dict:
 
 
 async def _run_docker_retention(session, now: datetime) -> dict | None:
-    """Drive the docker plugin's rollup + prune and its log ring via its
+    """Drive the docker plugin's event prune and its log ring via its
     capabilities, if loaded.
 
     Windows are read fresh from settings each run (rule 25 — a change in the
     Settings UI takes effect on the next hourly cleanup, no restart). Each runs
     in its own transaction, so a docker-side failure can't roll back the generic
-    prune above, and the log ring's deletes don't wait on the metrics rollup
+    prune above, and the log ring's deletes don't wait on the event prune
     (#5449). No-op when the docker plugin is disabled (capability absent).
     """
     from steward.core.capabilities import has_capability, invoke_capability
@@ -160,13 +159,10 @@ async def _run_docker_retention(session, now: datetime) -> dict | None:
     # Reads + rollup/prune share one transaction — get_setting's SELECT would
     # otherwise autobegin one, making a later session.begin() raise.
     async with session.begin():
-        raw_days = int(await get_setting(session, "docker.retention.metrics_raw_days") or 7)
-        rollup_days = int(await get_setting(session, "docker.retention.metrics_rollup_days") or 90)
         events_days = int(await get_setting(session, "docker.retention.events_days") or 30)
         counts = await invoke_capability(
             "docker.run_retention", UserRole.viewer, session,
-            events_days=events_days, metrics_raw_days=raw_days,
-            metrics_rollup_days=rollup_days, now=now,
+            events_days=events_days, now=now,
         )
     async with session.begin():
         logs_days = int(await get_setting(session, "docker.logs.retention_days") or 3)

steward/core/formatting.py

@@ -311,6 +311,7 @@ METRICS: dict[str, tuple[str, str]] = {
     "total_clients": ("Connected clients", "devices on the network"),
     "cpu_pct": ("CPU", "CPU in use; for a container, 100% is one core"),
     "mem_pct": ("Memory (% of limit)", "memory in use, against the container's limit or the host's RAM"),
+    "mem_usage_bytes": ("Memory used", "a container's memory in use, in bytes"),
     "restart_count": ("Restarts", "times the container has restarted, in total"),
     "disk_total_bytes": ("Docker disk use", "everything Docker keeps on the host (resource: the host)"),
     "disk_images_bytes": ("Docker images", "space Docker's images take on the host"),

steward/core/settings.py

@@ -117,11 +117,8 @@ DEFAULTS: dict[str, Any] = {
     # Days an ended metric anomaly waits to be reviewed before it lapses:
     # it leaves the to-review list and is kept as not reviewed (#5740).
     "anomalies.review_days": 14,
-    # Docker time-series retention (rule 25 — tunable, no restart). Raw 30s
-    # samples are heavy, so keep a short raw window then roll up to hourly
-    # averages kept much longer; lifecycle events are light, keep a month.
-    "docker.retention.metrics_raw_days": 7,
-    "docker.retention.metrics_rollup_days": 90,
+    # Docker lifecycle events are light; keep a month (rule 25 — tunable, no
+    # restart). Container CPU and memory follow metrics.retention.* (#5752).
     "docker.retention.events_days": 30,
     # Container logs (m79): on by default for every container (operator
     # preference). `exclude` names containers the server drops on ingest; the

steward/settings/routes.py

@@ -149,8 +149,6 @@ _THRESHOLD_FIELDS = [
 # Docker time-series retention windows (days). Read fresh by the cleanup task,
 # so a save takes effect on the next hourly run — no app.config wiring needed.
 _RETENTION_FIELDS = [
-    ("docker_metrics_raw_days", "docker.retention.metrics_raw_days"),
-    ("docker_metrics_rollup_days", "docker.retention.metrics_rollup_days"),
     ("docker_events_days", "docker.retention.events_days"),
     ("docker_logs_retention_days", "docker.logs.retention_days"),
     ("metrics_raw_days", "metrics.retention.raw_days"),

steward/templates/settings/thresholds.html

@@ -104,18 +104,13 @@
 </div>
 
 <div class="card" style="max-width:640px;margin-top:1.5rem;">
-  <h2 class="section-title" style="margin-bottom:0.5rem;">Docker data retention</h2>
+  <h2 class="section-title" style="margin-bottom:0.5rem;">Docker lifecycle events</h2>
   <p style="font-size:0.82rem;color:var(--text-muted);margin-bottom:1.25rem;">
-    Bounds how much Docker history is stored. Raw per-sample container metrics are
-    kept for the raw window, then rolled up into hourly averages kept for the
-    rollup window; lifecycle events are kept for the events window. Applied by the
-    hourly cleanup task — a change takes effect on its next run.
+    How long container start, stop, die and health changes are kept. Container CPU
+    and memory are kept with the other metrics, under Metric retention below.
+    Applied by the hourly cleanup task — a change takes effect on its next run.
   </p>
 
-  {{ days("Raw metrics", "docker_metrics_raw_days", "docker.retention.metrics_raw_days",
-          "Keep per-sample container CPU/memory points this long, then roll up to hourly.") }}
-  {{ days("Rolled-up metrics", "docker_metrics_rollup_days", "docker.retention.metrics_rollup_days",
-          "Keep the hourly-averaged series this long for multi-day history.") }}
   {{ days("Lifecycle events", "docker_events_days", "docker.retention.events_days",
           "Keep container start/stop/die/health history this long.") }}
 </div>
@@ -169,16 +164,18 @@
 </div>
 
 <div class="card" style="max-width:640px;margin-top:1rem;">
-  <h2 class="section-title" style="margin-bottom:0.5rem;">Host metrics retention</h2>
+  <h2 class="section-title" style="margin-bottom:0.5rem;">Metric retention</h2>
   <p style="font-size:0.82rem;color:var(--text-muted);margin-bottom:1.25rem;">
-    Bounds how much host-agent metric history is stored (CPU, memory, disk, network,
-    temps, …). Raw per-sample points are kept for the raw window, then rolled up
-    into hourly averages kept for the rollup window. Host charts read raw for the
-    recent part of a range and hourly for older data. Applied by the hourly cleanup.
+    Bounds how much metric history is stored: hosts (CPU, memory, disk, network,
+    temps, …), containers (CPU and memory), monitors and the other sources. Raw
+    per-sample points are kept for the raw window, then rolled up into hourly
+    figures (average, peak, and the level held) kept for the rollup window. Charts
+    read raw for the recent part of a range and hourly for older data. Applied by
+    the hourly cleanup.
   </p>
 
   {{ days("Raw metrics", "metrics_raw_days", "metrics.retention.raw_days",
-          "Keep per-sample host metrics this long, then roll up to hourly averages.") }}
+          "Keep per-sample metrics this long, then roll up to hourly.") }}
   {{ days("Rolled-up metrics", "metrics_rollup_days", "metrics.retention.rollup_days",
           "Keep the hourly-averaged series this long for multi-week history.") }}
 </div>