18032a4 re-checked a "not back in service" job every minute with no end,
so any later swarm wobble on the host (radarr's failed task during the
downloads move, say) flipped maintenance_failed back to 1: five false
firings and ten emails for Docker02 overnight on 10-19. And the overview's
finding never learned the rule at all, so it still said Docker02's
maintenance failed three days on.
The job now keeps came_back_at, set the first time its host is found back.
maintenance.still_failed(job) is the one answer: the figure, the overview,
the cadence page's run rows and the host panel all ask it. A failed update
or a timed-out job still flags its host until a job succeeds.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Three 404s on a quiet router read 7.5% and were raised as an anomaly; most of
the 4xx anomalies on review stood for 3 to 7 errors in 15 minutes. A new
Settings → Thresholds minimum, traefik.threshold.min_4xx (default 20, more
than one a minute held for the window), applies to both the router's grade
on its page and to anomaly detection, as min_5xx does for 5xx.
learn_normal takes `fewest`: for a share gated on a per-second rate, the
fewest events (rate × window × share) it must stand for to be judged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The host page gains a "Who's sending" panel, and the MCP a tool, that
put the host's outbound (or inbound) traffic down to its containers and
NFS mounts: read in shared time buckets, so the host's busiest bucket
says who was sending in it. A swarm service's tasks add up under its
name; what none of them accounts for is the host itself.
Also fixes CI 9488: per-container net_rx_bps / net_tx_bps join the
docker alert catalog, so a rule can catch a container flooding the
uplink.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Docker02's outbound traffic tops out the internet connection and Steward
couldn't say who sends it: a container's counters were kept only as a
running total, and NFS traffic belongs to no container.
- Docker ingest turns each running container's cumulative network counters
into net_rx_bps / net_tx_bps on "<host>/<container>" (net_rates), counting
from the stored row so a restart of Steward loses nothing; a container
new, recreated or restarted has no rate until its next snapshot.
- Agent 1.25.0 reads the bytes sent and received per NFS mount from
mountstats (already parsed for the delays) as tx_bps / rx_bps; Steward
stores them as nfs_tx_bps / nfs_rx_bps on "<host>:<mount>".
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Docker02's Sunday job updated and rebooted, then failed only because
portal_app was short of its replica past the 30-minute settle. Its
maintenance_failed figure stayed 1, and the alert kept firing, until a
later job succeeds. A job failed for not coming back into service is
checked again each minute and reads 0 once the host is back; a failed
or timed-out update still reads 1 until a job succeeds.
Also: compose.deploy.yml's syslog comment still said devices are told
apart by address under Swarm (stale since #6096).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
During Docker02's maintenance reboot its Agent offline alert fired muted by
the job's window, then resolved just after the window closed: the resolution
was mailed for a firing nobody had been told of. A resolution whose firing
was held and never released is now recorded with a note, not sent.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A pattern (phrases, a severity, a source, a window) is counted every minute
into plugin_metrics as lines_matching_<key> per added source, zero included,
so an ordinary alert rule fires on a match and clears once the window passes.
Each count reads only the lines stored since the last, by line id; a new or
changed pattern reads its window once.
Patterns page with ready-made storage-trouble presets and "Alert on it",
which opens the rule form aimed at the count. Source names are kept unique,
and a rename moves its counts and the rules on them.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- /plugins/syslog/sources/<id>: the lines newest first, loading as you
scroll, filtered by severity and worse, app, text and time, with day
dividers and times to the second in the reader's zone.
- The Syslog page links each source and shows its lines in the last hour.
- A Syslog panel on a host's page and an SNMP device's: the newest lines of
the sources bound to it or at its address.
- MCP: list_syslog_sources and search_syslog (source, host, text, severity
and worse, app, range, paging by next_before).
- Shared: escape_like moves to steward/core/like.py and cut_middle to the
MCP shaping module, replacing four inline copies; |when takes seconds=.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- plugins/syslog: RFC 5424 and RFC 3164 parsing, RFC 6587 TCP framing.
- Lines are queued as they arrive and stored every 5 s through a Holding,
so a storage stall delays them rather than losing them.
- A new sender is an offered source, counted and not kept, until an admin
adds it on the Syslog page (name, host); it can be ignored instead.
- The listener follows the plugin's settings (port, UDP, TCP), rebinding
within seconds of a save; a port it can't open is said on the page.
- Hourly ring per source: age window and byte cap from Settings.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Saving an enabled plugin threw away hot_reload_plugin's (ok, message), so a
plugin that failed to load live read "Enabled, not loaded" with nothing
more. Now:
- a live-load failure is recorded with its reason, as a boot failure is, and
logged; a later success clears it;
- any save of an enabled plugin that isn't running tries to start it, not
only the off-to-on toggle;
- enabling one already running is a success, not a failure;
- the plugin's page offers "Restart to load" while it is enabled and not
running.
Quart never marks the app as having served its first request, so
register_blueprint after startup is allowed: hot-loading is not refused
wholesale, as #6087 suspected.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The agent sent Docker's working set (usage less inactive_file), which still
counts recently used file cache. A file-heavy container with no limit read
191 GB while its processes held 2.2 GB, and the host's own figure, which
already leaves cache out, disagreed with it.
Agent 1.23.0 sends mem_used_bytes / mem_used_pct (anon + shmem, so
Postgres's shared buffers count; v1: rss + shmem) and mem_cache_bytes
apart. The new names keep the two meanings apart in history and in learned
normals: docker_013 renames the columns, clears the old values and adds
mem_cache_bytes. The Docker page, swarm by-node sums, Busiest Containers,
container detail, normals and the MCP tools all read the new figure. File
cache is shown only on the container's page and in get_container.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Each sample carries each NFS mount's requests since the last one: how many
a second, their round trip and execute time (READ and WRITE on their own),
and how many were sent again or hit a major timeout. mountstats is the
kernel's own count and never touches the share, so it reads while one is
stuck. Steward stores them per mount and the slowest on the host as
nfs_exe_max_ms, for an "NFS slow" rule and the host page's NFS chart beside
mounts_stalled; each mount's bar says what a request took.
Agent 1.22.0 (which also carries the backlog order from 41fc807).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The same Holding as the SNMP poll, for the rest of Steward's own
collectors:
- Monitors: the list is read within 10 s or the last one read stands in;
a start the database doesn't take is kept in memory, so a monitor isn't
checked twice; each result is held and stored at the time it was
checked. A stall shows on the monitors as the checks it saw, not a gap.
- UniFi and Traefik: each poll's writes are held (hold_write) and run
oldest first in one transaction; every reading is stamped with its scrape.
- A held round the database refuses (as opposed to can't be reached) is
stored alone and let go, so it can't hold up every round after it.
The Prometheus scrape is left as it is: its one target, Fabled Archives,
keeps its own database on the same NFS.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
On 2026-10-17 the NFS under Postgres stalled for ~15 minutes. Two things
went wrong for the readings that would have explained it:
- The SNMP poll stored before it polled again, so while the database hung
it stopped polling. It now collects on schedule and hands each round to a
Holding (steward/core/holding.py): one writer stores what is held, oldest
first, each reading at the time it was collected. Bounded at two hours of
polls, with a 30-minute deadline on a store; the device list read falls
back to the last one read after 10 s.
- "Mount not answering" (10 checks in a row) went pending on Docker02 and
Docker03 and never fired: the agent sent each post's backlog with its
newest, healthy sample, and that sample ended every post's run. Agent
1.22.0 queues the new sample behind the backlog, so Steward receives a
host's samples in the order they were taken.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A redeploy's new process ticks at once, so its once-a-minute samples never
broke for 90 s and get_anomaly showed none of the day's restarts. The first
self-monitor tick in a process now records restart_gap_secs (time since the
last process's last sample), and steward_gaps shows each one as a restart.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
One 502 among about 70 requests read steward@swarm critical at 1.4%:
past the 30-request minimum and far past the 0.1% cutoff, yet one
error. A new Settings → Thresholds field, "Fewest server errors to
fail" (traefik.threshold.min_5xx, default 3), must also be reached in
the 15-minute window; the 0.1% cutoff still catches a busy router.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
get_anomaly gives each occurrence against what is usual at that hour, the
metric around the newest one with its normal, what else happened on the
resource's host around each occurrence (alert transitions, Ansible runs,
other anomalies there and a count elsewhere, plus plugins' own events
through a new mcp_events hook: the host agent's maintenance jobs and
downtime, Docker's container starts, deaths and OOM kills), and how earlier
anomalies on the same metric were judged.
get_metric_normal reads the learned slots and the level that would be
flagged. review_anomaly takes one id or several, all or nothing, with a
note. get_anomaly_settings and (admin) set_anomaly_settings read and change
the review window and each normal's floor, as Settings → Thresholds does.
metric_series.points takes an optional `until`; core.alerts gains
host_of_resource; floor_setting names its floor.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5834 "The host agent is always on, and the MCP sees a host's
maintenance", step 1 of milestone 520 "The MCP sees what Steward has
grown".
The host agent is a core system (operator, 2026-10-15: "it's very core
to the functionality of Steward as a whole and shouldn't be considered
something that can be toggled off"). ALWAYS_ON_PLUGINS in core settings
makes to_plugins_cfg report it enabled whatever its stored row says, so
it always loads and its MCP tools are always listed. Settings -> Plugins
shows it as "Built in: always on" with no switch, and a save keeps it
enabled. Its code and URLs stay under plugins/host_agent: every
installed agent posts to /plugins/host_agent/ingest.
Maintenance through the MCP:
- list_maintenance_plans: each plan's group, weekly slot, reboot policy,
next slot, and its latest run (hosts, how far it got, who it skipped
and why, how it ended).
- get_maintenance_job(host | job): a host's jobs, newest and running
first, or one job: state, packages updated, reboot, drained by, why it
failed, when its current state times out.
- run_maintenance(host, reboot="if_required" | "never"): admin, as the
host page's Update now; refuses a busy host or one with no helper;
audited "<user> via MCP".
- The overview names a host whose latest job in the last week failed or
timed out.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5816 "Restoring mode": while a backup goes back in, the scheduler holds,
ingest and HTMX answer 503 with Retry-After (agents keep their samples),
and every page says what is happening. A safety backup of Steward as it
stands is made first, to /data/backups and the bucket. The backup's key
is put in place, the restore audited, and Steward exits with code 3 so
Swarm, compose or systemd starts it fresh. A failed in-place restore
changes nothing and says why on Backups; a full replace that fails after
the wipe restarts into first-run setup.
#5817 "Settings → Backups: Restore": a Restore button on each sent backup
and on the file card. The chosen backup is checked (the same comparison
as Check, now a shared partial) and held server-side for an hour under a
token; it starts only once the admin types "restore", and an older
backup also needs "Replace everything" ticked. A newer backup, or one
whose key can't open its secrets, gets no button.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5822 "Backup contents: settings and keys only". The operator's ruling
(System 180, 2026-10-15) overturns 2026-10-12's "Config + audit + alerts":
a backup is for system recovery and carries what would otherwise have to
be set up again, with no logs or history.
- contents.py: audit_events, alert_events, alert_states, metric_anomalies,
metric_anomaly_spans and user_invites move to EXCLUDED.
- Alert state is no longer restored, and evaluation skips a rule with no
state row. ensure_rule_states (alerts/service.py) gives every
single-resource rule lacking one an inactive row; alert_reconcile runs
it each minute.
- Docs, the Backups page and the CLI help say what a backup holds now.
CI's image check also finds the audit log empty after a restore.
- tests/integration/test_backup_restore_fixes.py is removed: its case (an
alert event naming a left-out Ansible run) no longer arises. The SET NULL
fix-up is covered by a unit test on synthetic tables.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Overview: a verdict line when something is wrong ('1 alert firing, 1
monitor down.'); its picker description now says what it lists (alerts,
monitors down, anomalies). Host figures have their home in Hosts.
Comments no longer name the Status page that was folded into Monitors.
- Monitors widget: a monitor not checked yet says when it was added.
- Host page: filesystems fullest first, by the worse of space and inodes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
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>
The root helper gains one read-only action, `drives`: it finds the drives
itself (smartctl --scan-open), reads each as JSON, and writes a summary the
agent carries on every sample (agent 1.20.0). Steward judges the figures in
drives.py (failing / worn / watch / ok / no SMART / unreadable), records them
per drive and as host roll-ups for covering rules, keeps the newest report on
the registration and lists it on the host page.
install-deps.sh installs smartmontools with apt, dnf, yum, zypper, pacman or
apk, run by install.sh and every agent playbook; one it can't install is
reported, never fatal.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
temp_c_max was recorded on every ingest, but host_agent didn't offer it
to rules, so "host running hot" couldn't be watched. The catalog guard
didn't see the row(...) helper ingest builds its rows with. Now offered:
temp_c_max, temp_c per sensor, and network and disk throughput per host,
interface and disk. A new guard holds every literal row(...) name to the
catalog, except NOT_ALERTABLE, where each exclusion is given a reason.
temp_c_max reads in °C.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5712 "Each router problem says what it means, whether it's real, and what
to do". A problem card now names its kind and reads its codes:
- Server errors: 500 means the app is failing; 502/503/504 means Traefik
couldn't reach it or it timed out.
- Client errors: 404 scanners, a 401/403 login wall, 429 rate limiting, or
499 clients giving up on a slow app.
- Slow: p95 against p50.
Each card says whether the problem is likely real, and ends with one next
step. A fault links to the swarm service behind the router, matched by
name, and traffic links to Settings → Thresholds. The wording lives in
plugins/traefik/meaning.py.
The plugin manager now hands out a loaded plugin's submodule
(loaded_plugin_submodule), so Traefik and host_agent reach Docker's
models the same way (#5775).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5768 "A maintenance cadence: a group's updates and reboots, one host at a
time, on a schedule". A plan names an inventory group, a weekday and time
in a time zone, a window and a reboot policy (only when required by
default). At its slot a run takes the group's hosts in order and gives each
a #5767 maintenance job. The next host starts only when the last is back in
service: its agent is posting and, for a swarm node, the node is active and
every replicated service is at its replicas. A failure stops the run, and
no host is started once the window has passed. A host's latest failure
shows as host_agent/<host>/maintenance_failed, for an ordinary alert rule.
The page is Hosts → Maintenance.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Step 1 of milestone 514 "Updates and reboots on a cadence, carried out by
each host's agent…": #5765 "The agent reports pending updates and whether
a reboot is required".
- Agent 1.16.0: an hourly background check (apt-check, apt list, or dnf -C
with updateinfo and needs-restarting), each command bounded at 120 s, no
root and no network. Debian's reboot-required marker and its packages
are read every sample.
- Ingest records updates_pending, updates_security and reboot_required for
rules, and keeps the newest on the registration (host_agent_005_updates).
- The Agent card leads with "Reboot required since …", then the updates
waiting or "Up to date", with when the check ran. list_agents carries
them too.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5764 "The agent reads a container's whole log backlog into memory, on its
main loop, with no overall deadline". Docker03's 1.13.0 agent went silent
for a day at 2.1 GB and 46 h of CPU, then ignored SIGTERM.
- One shared reader, _docker_get: each answer has an overall deadline and a
byte cap, not only a per-recv timeout. A trickling or endless body is cut
off (15 s / 32 MiB for JSON, 5 s / 4 MiB for a log read).
- Container logs: since the cursor but only the newest 2,000 lines, all
within 10 s. A note says when older lines were skipped, and a container
whose answer is too big skips ahead to now instead of re-reading its
backlog every interval.
- The agent reports its own RSS and CPU; Steward records agent_rss_bytes
and agent_cpu_pct, so a runaway agent shows before it goes silent.
Agent 1.15.0.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5760 "One stalled NFS mount freezes the whole agent: statvfs has no
deadline, so the host goes silent while it still answers ping". The agent
read every mount inline in its only loop; a hard-mounted share whose server
stops blocks statvfs in the kernel with no timeout. Each mount is now read
in a daemon thread under one 5 s deadline. One that misses it is sent as
stalled, and isn't probed again while its last probe is still blocked. A
mount that errors (a stale handle) is left out instead of emptying the list.
Steward counts them as mounts_stalled (alertable: "an NFS share stopped"),
and skips the fullest-disk figure when every mount stalled.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5756 "The agent says when its host rebooted and when it lost contact, so
downtime is recorded as what it was". Agent 1.14.0 posts events with the
next post that gets through: started (with the host's boot time), stopping
(on SIGTERM, with whether systemd says the host is going down) and
lost_contact (from its first unanswered post to the one carrying it).
Steward keeps them as the host's downtime (host_downtime): reboot,
agent_stop, unclean (a boot after the agent was last heard that nobody
announced) or lost_contact. Retried posts apply once.
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 (agents_away beside
#5721's registry). Past the grace, agent offline fires.
The host's Agent card lists the last 7 days of downtime. Rows are pruned
at data.retention_days.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
VDNT-Util01's / ran out of inodes (956 free of 60.8M) with 274G of space
left; every file create failed and Steward's page read healthy at 66%.
- Agent 1.13.0 reports each mount's inodes (statvfs; a filesystem reporting
none, like btrfs, is left without).
- Ingest records inodes_used_pct per mount, and disk_used_pct_worst is the
fullest mount by space or inodes, so existing disk-full rules and the
overview see it.
- The host's health strip, the Hosts widget and the host glance grade / by
whichever is fuller and name inodes when they are ("Disk / 66% · inodes
100%", "Inodes on / ... past its critical threshold").
Hosts pick up the inode figures once their agent is updated to 1.13.0.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
VDNT-Docker03's silent agent read as three overview findings (agent offline,
agent stale, the disk rule's no-data) and every drop emailed the disk rule's
silence and return on top of agent offline.
- The host agent registers its heartbeat (seconds_since_seen) and the
sources it carries (host_agent, docker). While a host's agent is past its
offline window, its other rules' no_data is recorded "while its agent is
offline" and not emailed, nor is their quiet return; a return past a
threshold still fires.
- The overview folds those rules under the host's agent-offline finding
("also silent with it: ..."), and a stale agent whose agent-offline alert
is active is that alert's finding, not a second one.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
An agent's flushed backlog brings samples hours old. The resume was stamped
at the first old sample's time, inside the silence it ended, and silence was
still judged by the backlog's sample times, so the series was marked silent
again between posts and resumed a second time. A resumed series now counts as
heard at ingest time, its events from that batch are stamped no earlier, and
the rest of the backlog reads as late samples.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Host CPU, memory and load per core, container CPU and memory, and every
monitor's response time learn a normal per hour (milestone 513, System 182).
A reading past its fixed cutoff but within its normal reads healthy and says
what is usual at this hour; one well out of it keeps its grade and is a
metric anomaly. Disk, a monitor down and an agent offline keep their cutoffs.
- Each normal's floor, the smallest departure that counts, is a setting:
Settings -> Thresholds -> Normals (norms.floor.*). Defaults: host CPU 10
points, host memory 5, load per core 10, container CPU 10 points of one
core, container memory 5, monitor response 100 ms; Traefik's 2-point 4xx
and 50 ms p95 floors move there from constants. learn_normal takes a
floor function (floor_setting) read at each pass.
- learn_normal(parts=False): a host learns, its cores don't.
- Ingest records load_pct (load_1m / cores x 100), the figure pages grade.
- Shared grading: anomalies.against / graded, worst_reading(normals=),
hosts' host_grades, docker usage(thresholds=, norms=), monitors'
latency_grade. Applied to host vitals, the Hosts widget, MCP overview,
Docker usage, and the monitor list, widget and page.
- The monitor and container pages list their anomalies.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5741 "Traefik routers are graded against their own normal", milestone 513.
- A router's 4xx share and p95 learn their normal (learn_normal in the
plugin's setup). Once a router has 5 days of history, a reading over the
fixed cutoff but inside its normal is healthy, and the healthy table says
it is usual for that router. One out of its normal is a warning that says
what is usual, and a metric anomaly.
- The 5xx share is a failure: fixed cutoff always, whatever the history or
traffic (System 182's ruling).
- A 4xx share over fewer requests than Settings -> Thresholds' new "Fewest
requests to judge 4xx" (30 per 15 minutes) isn't graded on the page, and
isn't judged for anomalies.
- Core: learn_normal(gate=(metric, least)) judges a key only while its
resource's gate metric averages at least `least`; `least` may be a
function, so it follows a setting. learn_normal(since=...) learns only
from history after a moment: Traefik latency learns from #5755's
per-scrape figures only.
- The Traefik page lists the routers' metric anomalies of the last 30 days
(_anomaly_card.html). The dashboard widget keeps its why to problems.
- Thresholds and docs say when the cutoffs give way to a normal.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#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>
#5740 "A metric that departs from its normal is recorded as an anomaly to
review, and is findable after it ends", milestone 513.
- Every minute each learned key's last 15 minutes is judged against its
hour's normal: above the usual hold x1.25 and by the metric's floor opens
an occurrence, which ends when the level comes back.
- metric_anomalies holds one row per item to review; a repeat while it
still waits joins it as an occurrence (metric_anomaly_spans).
- An ended row nobody reviews within anomalies.review_days (14, Settings ->
Thresholds) lapses; an ongoing one never does.
- Learning skips every hour an occurrence covers unless its anomaly was
marked normal: no normal is learned without review.
- Alerts -> Anomalies lists those to review first, then the reviewed and
lapsed under day dividers; marking is operator/admin and audited.
- The Overview widget and get_overview warn of open ones and count those
to review; the host page shows its own; the weekly report lists the
week's and the lapsed; MCP list_anomalies and review_anomaly.
- Reviewed and lapsed rows are pruned at data.retention_days, and kept
in backups like alert history.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A metric registered with learn_normal() gets, for every resource reporting
it, a usual hold (the 90th percentile across past days of each hour's p95)
and a usual peak for each slot:
- an hour of the day once 5 days of history exist;
- an hour of the week once history reaches back 3 weeks, so weekly jobs are
learned too.
Each slot takes in the hours either side, so a job that runs a little late
stays normal. One SQL statement relearns everything hourly into
metric_norms, from the rollup plus the raw window not rolled up yet;
norms_at() reads the slot a moment falls in. Nothing is registered yet:
Traefik (#5741) and hosts (#5742) opt in.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
plugin_metrics_hourly gains value_p95 beside value_avg and value_max. A
two-minute backup spike raises the hour's peak and leaves its p95 low; an
hour spent at that level raises both. Learned normals (milestone 513) need
that difference to call a recurring spike usual without calling the same
level held at another hour usual too. Hours rolled up before this keep a
NULL p95 and read as their average.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5691 "Generic Prometheus dashboard widget: pick a target and a template",
step 3 of milestone 508 "Prometheus widget templates".
- /plugins/prometheus/widget draws one target with one template in widget
shape: the verdict (led by "isn't answering" when the target is down),
then each block with problems first, figures beside a share bar or the
last 24h of bars, and the healthy rest folding to "+N more healthy" past
six. A missing target or template, or one that no longer checks, says so.
- library.py: bundled templates from plugins/prometheus/widgets/*.yaml,
the picker's target and template lists, and the default title.
- widgets/fabled_archives.yaml: the first bundled template, against the
series Archives reports (refined in #5693).
- Dashboard: a widget setting may take `options_from` (read when the picker
opens) with an `empty` sentence; `default_title`; a Details link may name
settings ("/plugins/prometheus/targets/{target}"). Prometheus is grouped
with the integrations.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5673 "S3 client and the backup job: settings, schedule, history, metrics":
- steward/backup/s3.py: SigV4 over httpx, path-style; put and list only
(an append-only key is enough; retention is the bucket's). Signing is held
to AWS's published examples.
- steward/backup/job.py: a 15-minute tick. The daily backup goes at the hour
chosen on the reader's clock; a missed day is made up, and a failure is
retried an hour later. History of the last 30 runs, audit
(backup.made / backup.failed), and backup_age_secs / backup_bytes under
"Steward itself", resource backup, for alert rules.
- Settings → Backups: verdict first, then the bucket (secret sealed; a
blank box keeps it), the schedule, Test the bucket, Back up now, and the
recent runs.
#5675 "Docs: backing up to Archives, and restoring": docs/core/backups.md.
Also: the audit page now words the Prometheus target actions from #5668,
which showed as codes, and a guard fails on any logged action without
wording.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The UniFi password sat in plugin.unifi in plain text: SECRET_KEYS covers
whole settings, and a plugin's settings are one dict. Saving the plugin's
settings page also wiped it, since the password box is never filled back
in and its blank value was stored.
- A plugin field whose name holds password, secret or token is a
credential (is_plugin_secret). set_setting seals it inside the dict,
every read opens it, startup seals any plaintext left from before, and
the undecryptable-secrets banner names it as plugin.<name>.<field>.
- The settings page uses the same test for its password boxes, and a blank
box keeps the saved value.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A bundled prometheus plugin, on by default like the other protocol plugins.
Targets are rows edited on its page: address, bearer token (encrypted),
interval, keep/leave-out name patterns and a series cap. Steward does the
filtering; the endpoint serves everything (Archives' ruling).
- A 5 s tick scrapes due targets at once, each within a 30 s deadline, and
writes the outcome to its row and its series through record_metrics, so
alert rules can watch any of them.
- Gauges are stored as they are; counters (and a histogram's or summary's
_sum/_count) as <name>_per_sec rates from the second scrape. Buckets and
quantiles are skipped. A labelled series is stored under
"<target>:k=v,…". Each target also records is_up, scrape_ms, series_kept
and series_capped.
- Pages: targets not answering first, with why; a target's series grouped
by metric with their newest values; a chart per series; the editor with
Test, which scrapes once and says what would be stored.
- The counter-to-rate keeper moves to steward/core/counter_rates.py; SNMP's
RateKeeper now scales its rate on top of it.
- metric_value reads Prometheus's _seconds as a span and _per_sec as a rate.
- The catalog guard treats an empty list as a dynamic source (snmp,
prometheus), as the rule form does.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
steward/core/prometheus.py parses the 0.0.4 exposition into families keyed
by name, each with its declared TYPE, so a scraper can tell a counter from a
gauge. A histogram or summary owns its _bucket/_sum/_count samples. Label
values are unescaped, and commas or braces inside quotes stay part of the
value. NaN and +/-Inf are read, timestamps are ignored, and a malformed
line is skipped. Traefik's parse_prometheus is now a thin reshaping of it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The Alerts widget (/alerts/widget) is shaped like every widget: a verdict
("2 firing, 1 silent." / "Nothing firing: all 12 rules are quiet."), then
what is firing or silent with its reading now, what the rule fires on and
since when, then what someone acknowledged and who; a line when alert
email failed in the last 7 days or is retrying (outbox.alert_trouble),
linking admins to Settings -> System; the quiet rules counted, a covering
rule once. It sits beside the Overview in the recommended layout (6x4
each).
The Overview's "+N more" line pointed at the removed /status/; alerts now
link to /alerts/ and monitors to /monitors/.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5509 "Delivery you can see: each alert's email state in history, and
outgoing email on Settings -> System" (milestone 501).
- Alerts -> History: an alert's email reads from the outbox (#5506):
"gathering" with when it sends, "sending", "retrying" with the attempt,
next try and the server's last answer, "sent", or "failed" naming each
reader it didn't reach and why (outbox.delivery / email_states). Events
from before the outbox keep what they recorded. The Notified column
shows once an alert has a queued email.
- Settings -> System gets "Outgoing email": every email of the last 7
days not yet sent, failures first, then the longest retrying, with
recipient, kind, state and last error, and a count of what was sent.
Failures and anything still retrying after 5 minutes lead the page's
verdict (outbox.outgoing).
- get_alert_history (MCP) adds `email`: the same delivery state.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5508 "Every other email through the outbox: weekly report, invites,
password resets, and the test buttons" (milestone 501).
- outbox.send_each queues one email per reader (claimed, so the sender
doesn't take it too) and tries them at once, for a page that answers
with what happened; outbox.said words the result. attempt takes a cfg
for a test of the form's unsaved settings and returns the settled rows.
- Weekly report (scheduled and Send now) and invites: queued, tried at
once; one the server couldn't take yet is retried for an hour. An invite
that didn't go shows its link with the reason, as before.
- Password reset: queued in the reset link's own transaction; the sender
delivers it on its next tick.
- Settings -> Notifications test email and a rule's Send test: queued and
recorded with retry=False, tried once (System 156 ruling).
- Removed mail.send, notifications._send_email and _smtp_send_sync;
dispatch_notifications is webhook only. A guard asserts only the outbox
(and smtp_open itself) opens an SMTP connection, with a self-test.
Also fixes run 8941 (#5507's integration test): its Also-send-to
addresses were "@test", which is_real_address rightly drops; they are
example.com now.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5507 "Alert emails through the outbox, gathered for 2 minutes"
(milestone 501). Replaces the in-memory core/alert_burst.py.
- outbox.hold_alert: an alert joins each recipient's held alert email, or
opens one due 2 minutes later (System 156 ruling). A Postgres advisory
lock keeps two alerts at once from opening two windows for one reader.
- email_outbox_events.alert_json (migration 0043) keeps the alert as it
fired, so the email written when the window closes says what it said.
- write_held writes it then, in the reader's zone: the usual alert email
for one alert, the list for several (outbox.alert_words).
- after_settle records on each AlertEvent, once every email carrying it
has settled: email_sent, or email_error naming each reader it didn't
reach and the server's reason.
- A held email survives a restart and goes out when its window closes.
- A refusal of a message's one recipient no longer repeats the address.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5506 "Email outbox: a table of outgoing email and a sender that retries
until accepted, for up to an hour" (milestone 501).
- email_outbox (migration 0042) holds each message per recipient, its
words, its state (held / pending / sending / sent / failed), attempts and
the server's last answer. email_outbox_events links alert events for
#5507.
- core/outbox.py: enqueue; a sender task every 15 s claims what is due
(SKIP LOCKED, stale claims released), sends the batch over one
connection with no session held, and settles each row:
- accepted -> sent;
- a 5xx refusal -> failed at once with the server's words;
- no connection, a timeout, a 4xx -> retried after 30 s, 1, 2, 5, 10,
15 min, the last try at the hour, then failed;
- a test (retry=False) settles on its first answer.
It runs on startup, so a row left by a restart goes out at boot.
- notifications.smtp_open is the one place a connection is opened.
- mail.message takes the html part as optional.
- cleanup ages email_outbox rows on the data window.
Senders move onto it in #5507 and #5508.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5469 "Settings → System: table sizes and what the last cleanup removed".
- Storage: each table's size on disk (pg_total_relation_size, indexes and TOAST
included) and Postgres's row estimate (reltuples), largest first, the 15
largest by name and the rest summed, under the database's total size. Read
from the catalog only, never by counting rows.
- Last cleanup: when it ran, how long it took, and what each pass removed (each
aged table, run log files, the metric and Docker rollups, the log ring), or
"Nothing was past its window". run_cleanup now returns that summary and keeps
it as the cleanup.last_run setting, as reports.last_sent_at does; cleanup_secs
(#5467) is recorded from the same figure.
- A hint links to where each retention window is set.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
