#5335 "Emails Steward sends read as designed messages: an HTML alert email
and weekly report, with a plain-text twin".
- core/mail.py renders each email as multipart/alternative from
templates/email/: a table-and-inline-style HTML part and a plain-text
twin carrying the same facts. Colours are the design system's light
theme, resolved from theme.css; nothing is fetched when a mail opens.
- Alert email: state badge, a headline sentence ("Disk used on web01 is
94%."), rule, resource, reading, the condition in words, and the time in
the zone Notifications was saved in with UTC beside it; "View in
Steward" when Steward's address is set. Subject "[Steward] Firing: <rule>
— <resource> 94%". Alerts now carry their operator.
- Weekly report: monitors down, active alerts, the week's alert events,
uptime and busiest Traefik routers as small tables with links. Sent from
the configured From address; the "Generated … UTC" line and the stale
repo footer are gone.
- condition_text() states a condition from a rule's parts.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The last page of #5214's walk. Settings still described Steward as it was
before monitors were unified and SNMP was rebuilt, drew its tabs below the
breadcrumb, and showed the report schedule in UTC.
- Tabs (#5377): settings/_tabs.html is a settings_tabs() macro called in
each page's section_tabs block, above the trail, as Ansible's are. The
breadcrumb guard also fails any page that {% include %}s a tab strip.
- Report time (#5378): reports.schedule_tz keeps the zone of whoever saved
the schedule; check_and_send compares in it (report_due), and the form shows
the day and hour on the reader's clock (schedule_in) under their zone's
name. "Last sent" uses |when.
- Wording (#5379): the report contents list what the report holds (monitors,
not ping-enabled hosts); "Ping latency" is "Response time" and says what it
grades; plugin detail says enabling loads now and disabling needs a
restart; the monitor interval says it applies on the next check (the
scheduler reads it every pass); the dead list-config branch is gone.
- Traefik (#5380): the 5xx and 4xx cutoffs are settings
(traefik.threshold.error_5xx_pct 0.1, error_4xx_pct 1.0) in a "Traefik
routers" group on Thresholds while the plugin runs, read by router_state
through error_cutoffs(). The Traefik page has a Thresholds button for
admins, to /settings/thresholds/#traefik.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
"0035_drop_plugin_catalog_settings" is 33 characters, and
alembic_version.version_num is VARCHAR(32), so the upgrade failed with
"value too long" (CI run 8800, integration). Renamed to
0035_drop_plugin_catalog. tests/test_migration_ids.py now checks every core
and plugin revision id fits, in the unit lane, which never migrates.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
UPS-MIB upsSecondsOnBattery is zero whenever the UPS is on mains power, so
"On battery for 0s" read as a timer that had just started. The reading gets
its own unit, battery_seconds: zero says "No, on mains", anything else
"for <span>". The metric key stays on_battery_for, so stored history keeps.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
First-party plugins have shipped inside the image since a7a281c, and the
catalog's only repository (Steward-plugins) answers 404, so Steward polled a
dead URL every minute for nothing. The operator chose to remove the catalog
entirely (#5214 "Review: readable and fast on the live instance").
Removed:
- core/plugin_index.py and the plugin_catalog_refresh startup task;
- the plugins.index_url default, and migration 0035 deletes the stored
plugins.index_url / plugins.repositories rows;
- the Settings -> Plugins repository editor, catalog, refresh, install and
reload routes and their templates;
- the download/checksum/extract install path in plugin_manager, and the
PLUGIN_INSTALL_DIR config it wrote to;
- plugins/index.yaml, docs/plugins/index.yaml.example and the guide's
"Publishing to the Catalog" section (now "Shipping a Plugin").
Kept: enabling and configuring plugins, hot-reload on enable, restart, orphan
cleanup, and the external root (/data/plugins) for operator-mounted plugins.
tests/test_no_plugin_catalog.py guards the routes, default and task.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
SNMP is on by default, so the loader has already mapped snmp_devices under
_steward_plugin_snmp; importing plugins.snmp.models defined the table twice
(integration lane, run 8788).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The operator, reviewing SNMP on #5214: "this system looks incomplete like
there's something missing from it". The one device shown was plugin.yaml's
example (core-switch 192.168.1.1, community public), polled every minute;
devices could only be defined by editing YAML in the image; a failed poll was
logged at DEBUG and stored nowhere, so the page said "waiting for the next
poll" forever; v3 ran as v2c; counters showed as lifetime totals.
- Devices are rows (snmp_devices, migration snmp_001): address, port,
v1/v2c community or a v3 user with auth/priv protocols and keys (secrets
encrypted), presets plus custom readings, an optional host binding, and
the newest poll's outcome. The YAML device list and its example are gone.
- The SNMP page adds, edits and deletes them (admin, audited, with
get_audit_entities so the audit log names and links a device). Readings
come from presets (system uptime, interface traffic by number, UPS-MIB,
Linux net-snmp) and custom "OID name [unit]" lines.
- One GET per device with every OID, devices polled at once. The outcome is
written to the row in words ("No reply from … within 5 s. Check the address
and port, that SNMP is on, and the community"), with readings it lacks.
v1's whole-request error drops that OID and asks again. SNMPv3 signs in
with USM (MD5/SHA-1/SHA-2, DES/3DES/AES).
- Test asks the unsaved form's device for its name, description, uptime and
interfaces by number. It is an admin's explicit action; every page reads
only stored rows (System 178's ruling).
- Each reading has a unit; a counter is stored as a per-second rate (bytes as
bit/s), a reset or wrap skipped. The page, widget and host panel say values
in units and why a device is silent; the device page draws a chart per
reading on a dated axis.
Tests: readings and presets, the form and its secrets, v3 checks, rates,
the poller against a stand-in pysnmp (one GET, v1 retry, timeout wording,
USM), the pages, and an integration test of the poll writing outcomes and
rates on Postgres.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The operator, reviewing the audit log: "the dividers are good but should
stick out more maybe make the text color different". Column headers are
muted text on the card; a divider is now the primary text on the hover
surface, ruled above and below. The accent stays off it: the design system
keeps it off chrome, and Steward's amber sits near its warning colour.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The plugin_catalog_refresh tick retries a failed repository every 5 minutes;
the configured index URL answers 404, which logged a full traceback each time.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The operator's ruling (Background Jobs, 2026-10-08): a Steward page reads
stored records; the sources are contacted by scheduled collection. An audit
of every GET handler, HTMX fragment, widget poll and MCP read tool found one
page load that reached out: Settings -> Plugins' catalog fragment fetched
each plugin repository (5 min cache, 10 s timeout each, retried on every
load when unreachable).
plugin_catalog_refresh ticks every minute and fetches a repository whose
copy is missing or older than an hour, waiting 5 min after a failure;
saving repositories clears the cache, so the next tick fetches. The
fragment only reads the cache and says which repositories are not fetched
yet. Refresh is now a POST that fetches at once, as does install.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The phase log on the live instance: 'traefik rows: slow 7.4s (latest=7.04s
history=0.11s window=0.04s ...)'. latest_per_router walked traefik_metrics'
index one router at a time; each step lands on the router's oldest
entries, which retention deletes every minute (likely dead ones). The
scraper already records every router figure through MetricBatch, which
keeps one row per router and figure in plugin_metric_latest (#5208's
table, read as latest_by_resource does); latest_per_router reads that
now, and the router figures carry the scrape's own time.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The operator: "load was still slow" (the access log: 7.6-8.0 s on every
poll, warm), "why do the bandwidth and 4xx sections not also get
sparklines", and "no gap between the widgets". router_history reads the
4xx share and bandwidth too, and the card draws all five lines. The
problem cards sit in .node-grid with a bottom margin; a CSS column box
dropped the last card's margin. rows() and widget() log each phase when a
request passes 1 s, so the slow query names itself on the next load.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The operator: problem cards "can disappear so quickly". Each router was
graded on its latest scrape, so a 4xx share near the 1% cutoff crossed it
on one scrape and fell back on the next. recent_window reads every
router's last 15 minutes in one query, the error shares weighted by
traffic; grade_routers uses it on the page and the dashboard widget, and
the why says "over the last 15 min".
traefik_config_last_reload_success is the Unix time of the last good
reload, not a 0/1 flag, so the tile said "Reload failed" on every scrape.
traefik_005 stores it as config_last_reload_at, and the tile reads
"Config reloaded 3m ago". The router index also carries the 4xx share.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The operator found "Today 08:03" in place of a date odd on the audit log,
and asked for the recency groups as dividers instead. |when now always
writes the date ("Thu 8 Oct 08:03"); mark_day_groups gives each entry its
group in the reader's zone, and a divider row opens each group. A scrolled
page carries the group the last one ended in, so it is not opened twice.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5355 "Traefik page doesn't load": /plugins/traefik/rows took 12-28 s.
It ran a query per router ever seen, and each raw row in the range was a
heap fetch. traefik_004 rebuilds the router index to cover the charted
columns and indexes scraped_at for the request chart; router_history reads
the current routers' buckets in one query.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Agent 1.12.0 stamps each post with sent_at; ingest compares it with the
moment the post arrived: clock_offset_s / clock_checked_at on the
registration (host_agent_003), and a clock_off_secs metric (either way)
that alert rules can watch. Replaces the old skew log, which measured the
newest sample and was off by up to an interval.
- Settings -> Thresholds: Clock drift, 30 s warn / 5 min critical.
- Host page: 'clock 0.3s ahead' in the vitals line, toned past the warning.
- Hosts list: names hosts whose clock is off; when every reporting host is
off by about the same amount, says Steward's own clock is the likely one.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Run 8736: test_user_model_fields listed the columns without time_zone, and
the session-refresh fixture's users had no time_zone attribute. The refresh
test now also asserts a zone set on the Account page reaches the session.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- users.time_zone (0034). An account without one takes the browser's zone
on its first page (POST /account/time-zone/adopt, never over a choice made
by hand); the session follows the account on every request.
- /account: who you are and the zone every page writes its times in,
reached from your name in the sidebar.
- formatting.when: 'Today 06:58', 'Yesterday 23:41', 'Tue 6 Oct 14:12',
'6 Oct 2025 14:12', 24-hour, in the reader's zone; |ago's hover is the
exact local moment instead of UTC; tz_label() for column headers.
- Audit log: When (EDT) with exact times, one detail fact a line, on classes.
- tzdata in the image, and the airgapped smoke checks a zone resolves.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The tabs were drawn in each page's content block, which base.html renders
after the breadcrumb, so on every nested Ansible page the strip moved down a
line. base.html now has a section_tabs block above the breadcrumb, and every
Ansible page fills it. A guard in test_breadcrumbs.py fails any page that calls
a tabs macro outside that block.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Every run warned twice because of Steward's own inventory:
- 'Invalid characters were found in group names': the group 'Docker
Swarm' went into the inventory as typed. ansible_group_name() cleans it
to docker_swarm (lower case, [a-z0-9_], never a leading digit); two
names that clean alike get _2, _3 so neither folds into the other. The
group page says what plays call it. The display name is unchanged.
- Python interpreter discovery: ansible_env sets
ANSIBLE_PYTHON_INTERPRETER=auto_silent unless the environment already
sets one; a target or group's ansible_python_interpreter still wins.
Runs are scoped by target set (steward:group:<id>), not by group name,
and --limit is free text, so neither path depends on the name.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
/ansible/inventory/ shows every target (address, its linked Steward host,
its groups) and every group (members, the schedules aimed at it) under a
verdict line. The whole row opens the target or group. Add target and
Add group open small dialogs; a failed add comes back with its dialog
open and what was typed. The red x leaves the rows: Delete is on the
target's or group's page, behind the confirm dialog, saying which
schedules would be left with nothing to run on.
The two list pages (/inventory/targets, /inventory/groups) are gone
with no redirect (rule 122); their URLs keep only the Add POST.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
CI run 8705 on 0c72892: the schedule page's no-inline-style guard caught
_playbook_vars.html (shared by the run and schedule forms), and the run
page's Run again assertion left out the link's class.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- The Schedules tab: a verdict line (failing first, paused, next due),
each schedule as one line (what runs where, how often, when next, how it
last went); the row opens its page; Run now and Pause stay in the row.
- /ansible/schedules/new and /ansible/schedules/<id>: the form on classes,
dry run beside the button, the playbook's variables rendered with the
page; at the foot Delete behind the confirm dialog. ?source=&playbook=
start a new one on a playbook (Schedule it on the playbook's page).
- A run started by a schedule links to that schedule's page.
- The run page's id guard allows the Run again link's ?from=.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- /ansible/playbooks lists every playbook across sources: what it says it
does, its category and destructive flag, how it last ran; the whole row
opens the playbook. A source with none is one quiet line.
- /ansible/playbooks/<source>/<path>: the verdict (what it does, flagged
destructive, last ran), the run form in place (dry run beside the button,
options folded), the schedules that use it, its recent runs, the playbook
itself folded. ?from=<run> fills the form from that run: Run again on a
run's page.
- /ansible/browse, its view page and the /run-form fragment are gone, not
redirected (rule 122); every link points at the new pages.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- section_tabs macro; Runs · Playbooks · Schedules · Inventory on every
Ansible page, replacing each page's own set of header buttons.
- Runs: a verdict line (last 30 days, failures first with the latest named,
running now, the schedule due next); under each status what the run did
across its hosts (ansible/recap.py); a filter bar (playbook, status, who
started it) that swaps the table and carries into the scroll paging.
- Run page: status, duration, start and who started it in one line; each
host's result in words, worst first; the playbook it ran links to it;
the log and failures on classes rather than inline colours.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The name's link stretches over its row (.row-link / .row-target); the
History and Disable buttons sit above it and keep their own clicks.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The all-rules page reads a Postgres shared with other tests, whose
Ansible-run events turn the column on.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
History (#5316):
- Transitions fold into episodes (steward/alerts/episodes.py): each stretch
out of quiet, the worst state it reached, how long it lasted, its worst
reading against the rule ('3m 21s (fires above 3m)'), who was told across
its changes, and its Ansible runs. The changes themselves fold under it.
- An episode cut by paging says so ('cleared later (above)', 'earlier').
- For one rule: a verdict ('Fired 4 times in the last 7d, 1h 2m in all') and
no Rule column. Notified and Action show only when the range has any.
Tests:
- The confirm guard looks for a confirm( call, not the word in the comment
that says it is gone (lesson #4365).
- The Hosts viewer test looks for the add-host dialog by id, since every page
now carries the confirm dialog.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Confirm dialog (#5314):
- Every data-confirm form now asks in one native <dialog> (base.html + static/confirm.js)
instead of window.confirm.
- The question becomes the heading and what follows it the detail; the act's
button is named after the form's own button; Cancel has the focus and Esc
backs out. Optional data-confirm-title/-detail/-action/-dismiss attributes.
- All 27 existing uses work as they are. Cancelling a run says 'Keep running'.
Alerts list (#5315):
- A verdict line (what is firing, how many rules, where notifications go,
or 'notifies nobody yet' with a link) replaces the empty Active card.
- Each rule's name opens it; rows show 'last fired' and fold fire/clear into
the condition; History and Disable stay; Delete leaves the list.
- .docker-verdict becomes the shared .page-verdict.
Rule page (#5317):
- Named for the rule (title and crumb), with when it last fired and how many
changes of state it holds.
- Three sections: what it watches, when it fires (checks side by side,
seconds echoed as a duration), what happens (where notifications go, no-data,
the playbook, whose fields stay folded until ticked).
- Delete at the foot, behind the dialog, saying its history goes with it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Service page:
- The header carries the verdict, mode, stack, image and when it was reported.
'In trouble since' shows only while the service is in trouble; once recovered,
'Last failed X ago · its logs'.
- The failure card shows only while the service is down, degraded or crash-looping.
- The tile strip and the kv card are gone.
- One Tasks table: running tasks first with CPU, memory and since when, then
Swarm's record. The node is named as its Steward host throughout, and
'N failed in 24h' sits in the caption.
Container page:
- Same shape: header verdict, a 'Last exit' card only while it is stopped
after a failing exit, set-up and I/O in one kv card, and lifecycle as a table.
- A swarm task's breadcrumb goes through its service.
Both pages:
- Resource history is the one fragment (_usage_history.html): now and peak in
cores and bytes, drawn with a floor of 1 core / 1 GiB (sparkline_svg floor=),
so idle noise draws flat.
- The container page's old % sparklines, its private _sparkline and the unused
bucket()/container_history() helpers are removed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- #5304 "Docker disk totals recorded over time per host, with a growth chart and alertable metrics"
- #5305 "Disk page: images grouped by repository, versions folded under it"
- #5306 "Docker overview: one disk line per host linking to the Disk page"
Each `docker system df` report is now also recorded as metrics, with the host
as the resource: disk_total_bytes, its parts, and disk_reclaimable_bytes. An
alert rule can watch them. The Disk page starts with the hosts side by side,
showing how much each grew over the selected range with a line of it, then
gives each host its own section. An image's versions fold under its
repository, which is charged only for the layers its versions don't share.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5303 "Swarm page laid out by node: role, resources and the tasks placed on each"
A verdict line first, then the nodes that are down or drained and the
services short of replicas, then one card per node. A node with a Steward
agent shows what its swarm tasks use against the machine's cores and memory,
service by service. A node without one lists Swarm's placement, by count.
The services table moved to the Docker overview, where it can be filtered.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
render_plugin_template's first three arguments are positional-only, so the
service page's own `name` reaches the template. The overview names a service
rather than its replica containers, and a stack chip keeps the search it was
clicked under.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- #5298 "Docker overview: verdict line; only the problem tiles that aren't zero"
- #5299 "Docker overview: swarm services as a filterable, sortable table with a Stack column"
- #5300 "Docker overview: failures listed once, each with a link to the failed task's logs"
- #5301 "Docker overview: standalone containers in one table with a Host column"
- #5302 "Service page: replicas, image, task history, resource history and logs across all its tasks"
overview.py holds the rows, filters and sort as plain functions. The filter
bar sits outside the polled fragment and the URL keeps its state. A service's
page reads every task container Steward still holds anything for, the live
ones and removed ones known by their events, so a failed task's logs are
reachable after Docker has removed it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
CI 8616: the integration test asserted Vary == 'HX-Request', but the session
layer appends 'Cookie'. resp.vary.add keeps both; the test asserts membership.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5294 "Hosts list: search, group and status filters; sortable columns; a Groups column".
Ansible inventory groups are Steward's grouping system (operator's ruling): a host's
groups are its linked target's, and membership overlaps, so the list filters by group
rather than sectioning by it.
- Filter bar: search on name or address, a group (or "No group"), a status (Down,
Agent silent, Up, No monitors). The table swaps in alone and the URL keeps the filters.
- Name, Status, Latency, 30d uptime, CPU and Memory headers sort (sort_th in _macros);
problems-first stays the default, and a host with no value goes last either way.
- A Groups column of chips, each filtering to its group; shown when any group exists.
- A swap is keyed on HX-Target, so Back to a pushed URL still gets the page.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The stylesheet's reset (* { margin: 0 }) removed the browser's margin: auto
that centres a modal <dialog>; .modal puts it back.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5290 "Host page: the health tiles span the card's width; even spacing above Ansible":
the four tiles take equal quarters (halves on a phone) and their sparklines stretch
across them; the Monitors + Agent row drops the Agent card's own margin, so the gap
above Ansible is 1rem like every other.
#5291 "Add host opens as a dialog over the Hosts list": a native <dialog> on /hosts/,
opened by Add host or /hosts/?add=1. A name clash re-renders the list with it open,
the error and what was typed. GET /hosts/new is gone; hosts/form.html is Edit only.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Hosts list:
- CPU and Memory are their own labelled columns (#5279); a silent
agent's last readings are greyed, and Status says "agent silent for 7m".
- One uptime figure, 30 days, with 24h and 7d on hover. The SLA tab
(/hosts/uptime) repeated it and is removed (#5280; one home per fact).
- Status is a word with how long ("Down for 12m"); host_status now carries
down_since. Edit and Delete live on the host's page only (#5281).
- Figures are right-aligned with tabular digits (.table .num), not a
monospace face.
Agent fleet (#5282): its add-host form is gone, so hosts are added only on
Hosts -> Add host. That form was also the only way to get a token for a
manual curl install, so that moves to the host's Agent card as "Install
manually" (POST /fleet/<host_id>/register), shown on the fleet page as
before. "Delete" there reads "Unregister".
Add host (#5283): laid out like the add-monitor form; a name clash comes
back inline with what was typed instead of a bare text 400; the address
is optional, as the host page already allowed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Docker panel (#5276): problems first, then the 8 busiest running
containers by CPU; the other running and the stopped fold behind one
"+N more running · M stopped" line. Containers that are not swarm tasks
read the way the Docker page reads them: CI jobs as "CI: workflow · job",
buildx builders as "BuildKit builder · name".
Monitors card (#5277): the edit/delete icon buttons and the inline add
form are gone. Each row links to its monitor's page; "Add monitor" opens
the add form filed under this host, starting from its address, with the
host in its breadcrumbs and Cancel back to it.
Agent card (#5278): Full metrics stays in view; Update, Rotate token,
Unregister agent (was "Remove") and Re-provision sit under "Manage agent".
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
DNS monitors take an optional Resolver, a server's IP. Set, the check asks
that server directly for A and AAAA over UDP and records how long it took
to answer; empty keeps today's lookup through the container's resolver.
"Does my local DNS still answer for public names" and "does 1.1.1.1" are
now separate monitors. No new dependency.
On a host's page the Ansible card moves above the Docker list, which can
run to dozens of containers and hid it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Monitor page: a response-time chart for the last day with failed checks
marked; Edit / Pause / Delete as labelled buttons; the state tile says
"for 29m" instead of "for at least 29m 25s ago"; the check strip fills
its card; no failures is one line, not an empty box; Settings names the
type, target and the actual global interval, with readable field names.
Add monitor: a line under the fields says what the selected type checks
and when it counts as up; Name says it defaults to the target.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Status repeated the Monitors page's counts, check strip and uptime. Monitors
now has one header row and every column Status had: the last 30 checks,
latest response with its trend, uptime over 24h / 7d / 30d (the range picker
goes) and TLS days on every row. /status and /status/rows and the nav link
are removed outright; the dashboard Overview (/status/widget) stays and its
Details link goes to Monitors.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A chart built while its cell was still being laid out could stay blank until
a window resize; it now measures and draws again after layout settles. Each
refresh rebuilds the chart (the flash), and its buckets are minutes wide, so
the widget refreshes every 5 minutes instead of every poll.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Each figure after its bars takes a fixed width, so memory figures of
different lengths no longer shift their bars. CPU and Memory headers say
"last 6h" and the note gives the slice length.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The fill bars were a share of the host's capacity, a sliver on hosts with
80 cores. Each figure now sits beside 24 bars of the last 6 hours (1h/24h
per widget), each row scaled to its own peak but never below 1 core or
1 GiB (configurable per widget), so a 0.05-core container reads flat and
a bar's height means real load. One grouped date_bin query feeds them.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
View mode splits each saved row into seven, so a widget no longer carries up
to 69px of empty space below its content (Overview, SNMP). Edit mode and the
saved layout keep 70px rows. The Busiest Containers host column aligns left
like its header, its note fits one line, and the Hosts column reads Disk.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
CI 8551 integration: "integer out of range" for a 2099 timestamp.
The SQL in the log showed a second problem: `CAST(epoch AS INTEGER) / 60`
compiled to `/ CAST(60 AS NUMERIC)`, i.e. decimal division. Every scrape got
its own "bucket", so the request chart and the Traefik page's sparklines
never averaged anything. The int4 cast would also overflow in 2038.
Both now use date_bin from the epoch, as host_agent/metrics_query.py does.
The integration test adds a second scrape 30 s later in the same minute and
checks that the bucket averages the totals and weights the 5xx share over
both scrapes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5256 "Traefik — Request chart: blank on the live dashboard; time axis and honest labels".
- "Req/s" was the average of the routers, not the total. Each scrape is now
summed over routers, then averaged per bucket (request_history).
- The 5xx share is now weighted by each router's traffic. The axis is no
longer pinned at 0–100 (suggestedMax 5).
- The x axis shows times, in the viewer's local time. The tooltip shows
both series at that time.
- A verdict line comes first: now / peak req/s and the peak 5xx share. The
figures read even if the chart doesn't.
- The blank live chart's cause is still unknown. Chart.js and
chart-colors.js both load on the dashboard. A missing library or a thrown
error is now written into the widget instead of leaving it empty, so the
cause will show on the page.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5255 "SNMP — Devices: status in words and key readings".
- Verdict first: "All N devices answering", "ups: no reply since 14m ago",
"N of M devices not answering", or "N not polled yet".
- Devices that stopped answering come first, then those never polled, then
the rest.
- One header row: Device / Readings / Polled.
- The name is the link; the separate "detail →" is gone.
- The hollow dot is replaced by words: "no data yet", or "no reply since …;
readings are from then".
- Key readings: the first two in config order, gauges ahead of counters.
Units come from the label suffix: _centisecs as a duration, _pct, _dV as
volts, _octets as bytes total.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
CI 8548: '<th' also matched '<thead', and the certificate test's clock was
read before its certificates were built, so 10 days came out as 10.00…
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5254 "Traefik — Routers: current routers only, header row, certs on the router they serve".
- Router hosts come from Traefik's API (/api/http|tcp/routers), read
best-effort on each scrape. An optional api_url overrides it; the default
is /api on the metrics host. A certificate's expiry then shows in a TLS
column on the router whose host it covers. An expiring one makes that
router a problem, graded by Settings → Thresholds.
- Only current: routers and certificates must have appeared in the last
three scrapes. Certificates are upserted by serial and never deleted, so
renewed or retired ones lingered on the widget.
- Widget order: verdict, problem routers with the reason, then the busiest
healthy routers under one header (Router / Req/s / Bandwidth / p95 / TLS).
A missing value shows "—".
- If the API is unreachable, expiring certificates are listed after the
routers, not above them.
- The page splits certificates into served (with the routers using each,
plus SANs) and "No longer served".
- Latest row per router: an index walk (recursive CTE + LATERAL) replaces
DISTINCT plus a query per router.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The widget names the busiest container in a verdict ("Busiest: tdarr on
VDNT-Docker03, 1.6 of 8 cores."), then a fixed-layout table (container,
host, CPU in cores, memory in bytes) with each figure's share drawn as
shading behind it: of the host's cores, and of the memory limit or the
host's memory.
The numbers were already rendered, but each row grew as wide as its longest
name. A grid item won't shrink below its content, so Gitea CI container
names pushed the figures out of view and the cell scrolled sideways. Names
now come from display_name(): a swarm task by its service (no slot number),
an Actions job container as "CI: <workflow> · <job>". They cut off with …
and show whole on hover.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A verdict line (services healthy · other containers running · failed in
24h), then "Needs a look": troubled services with why (crash-looping, N of
M running), since when and the last error, and troubled containers
(unhealthy, restarting, exited with an error, out of memory). Then "Failed
in the last 24h" by name, with how often, when and where. Then the healthy
rest one line each (name · hosts · replicas) in as many columns as the
widget is wide, collapsing after 40 to "+N more healthy".
Per-container CPU is gone from this widget: Busiest Containers owns it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The Hosts — Overview widget opens with a verdict ("VDNT-Util01: CPU high."
or "All 3 hosts healthy."), then one row per host under one header: CPU,
memory, disk / and load. Rows sort by host_glance: a down monitor, a
reading past critical, a quiet agent, a warning, then the healthy. Load
reads "16.8 / 8 cores", graded per core as Settings → Thresholds intends
(cores counted from the per-core CPU sub-resources, as on the host page).
A quiet agent's last figures are shown uncoloured, with when it last reported.
Ping and 24h uptime are gone from this widget: they belong to Monitors. So
are the unlabelled CPU sparkline and the 30-day uptime query it no longer
needs.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The Monitors widget opens with a verdict ("1 of 5 monitors down."), then one
row per monitor under a single header: latency and uptime over 24h, 7d and
30d, down monitors first with why and since when, the name linking to the
monitor. Counts follow the type filter. Problems always show; the limit
trims only the healthy rest ("+N more up").
Uptime / SLA (uptime_summary) repeated those numbers per host; migration
0033 moves dashboards onto Monitors. Shared widget styles (.w-verdict,
.w-table, .w-sub, .w-more) carry the widget-shape ruling for the widgets
that follow.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Each widget is as tall as its content, never taller than its saved height
(where it scrolls). The dashboard is capped at 2400px instead of 1400px.
Past 12 columns' worth of width it adds columns (one per 110px, up to 24)
and places widgets again in reading order at their saved widths, so more sit
side by side instead of stretching. Edit mode shows and saves the
12-column layout, so a layout doesn't depend on the screen it was edited on.
The share view uses the same layout code (static/dash-layout.js).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- New installs start from widgets.RECOMMENDED, packed row by row for the
enabled plugins; the edit drawer can reset a dashboard to it.
- Host Agent — Resources folds into Hosts — Overview (which gains load 1m),
UniFi — Devices into UniFi — Overview; migration 0032 moves existing
dashboards onto the survivors.
- The Overview labels what it counts (hosts, monitors up/down/pending,
alerts firing) and lists firing alerts and down monitors with since-times.
- Uptime polls every 5 minutes (poll_seconds) instead of every cycle.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5212 "Plain labels: no raw ids, scopes, metric keys or unexplained jargon",
batch 3:
- Disk page: each tile says what it holds. "Writable layers" becomes
"Container files" and Build cache says it is safe to clear. Shared is
"Shared with other images" with a tooltip, and "dangling" is gone.
Untagged images show by id instead of "<none>", a capped list says
"The 50 largest of N images", and each host shows "as of" its report.
- Traefik: an expired certificate reads "expired 12 days ago", not "-12
days" (cert_left). The Traffic empty state knows whether the access log
is on: off points to Settings → Plugins → Traefik; on says to check the
mounted log path.
- Moments use |ago with an exact UTC tooltip: UniFi last seen, alarms,
events, speed test and DPI; AI token last used.
- Settings: the Auth tab is "Authentication" like its breadcrumb, plugin
fields read "Metrics URL" not metrics_url (field_label), a source's pull
interval is "every 1h", and the last report time drops the ISO "T".
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5212 "Plain labels: no raw ids, scopes, metric keys or unexplained jargon",
batch 2:
- A run records where it came from: AnsibleRun.origin ("schedule", "alert"
or "mcp") and origin_id, set by the scheduler, Run now, alert actions and
the MCP tools. Migration 0031 backfills past runs from alert events,
schedules' last runs, and MCP audit entries.
- The runs list has a Started by column: "schedule nightly", "alert rule
nas down", "alice via MCP" or "alice", linked where there is a page.
The raw scope tooltip is gone.
- Run detail: no run id in the subtitle or body, Started by in words, and
the host summary lists failed and unreachable hosts first. The live
output only follows new lines while the reader is at the bottom, and
when the run finishes the page reloads once so the badge and summary
are final.
- list_ansible_runs and get_ansible_run report origin and origin_id.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Run 8504 failed unit: tests/core/test_alert_history_row.py built a rule
with no metric_name, which _history_row now reads to say the reading in
words. The rule now names cpu_pct, and the value and condition asserts
expect "CPU 98%", "No CPU reading" and "CPU > 90%", plus an up/down case.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5212 "Plain labels: no raw ids, scopes, metric keys or unexplained jargon",
batch 1:
- Audit log: each action reads as a sentence ("paused monitor"), with the
code as its tooltip. Detail shows flags as yes/no, lists joined, a run's
scope by name, spans as spans, and no ids the entity column already gives.
The filter is now a picker of areas.
- Alert rules, history and Status say the condition in words: "is down",
"Fullest disk >= 85%". Readings use metric_value, and history times use
|ago.
- Rule form: sources and metrics by name, the threshold's unit, "fires
when it is down/up" for an up/down metric instead of 0.5, and a metric
reference built from the catalog.
- Schedules and Maintenance: viewers no longer see Run now, Pause, Edit,
Delete or New. Intervals and next runs read as "every 2h" and "in 3h";
a scope reads "Every alert on Host agent".
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5211 "Lists load as you scroll: audit, alert history, Ansible runs"
(milestone 477), per preference 172.
- core/paging.py: keyset paging over (time, id), newest first, with an
opaque "<iso>|<id>" cursor; a page reads one row past itself to know
whether another follows. No OFFSET, so rows added while someone scrolls
never shift the pages after them.
- _macros.html scroll_more / scroll_end and static/scroll-more.js, written
once: an IntersectionObserver with a 600px margin fetches the next page
before the reader reaches the end; "Loading more…" while in flight,
"End of the list." at the end, and a failed page says so with a Retry
button instead of re-firing on every scroll.
- Audit log: no "Load more" link and no 500-row cap.
- Alert history: no 200-row stop with "pick a shorter range";
alert_svc.history takes the cursor, so the MCP tool and the page share
one query.
- Ansible runs: no silent cap of 50.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5210 "Problems first, with since-time, across every list" (milestone 477),
second batch.
- Host page: an "Alerts on this host" strip first (rules on its agent
series, its containers or its monitors: alert_svc.watches_host); its
monitors problems first with "Down 12m · reason"; Docker above Ansible.
- Uptime page and widget: worst 24h uptime first, then 7d and 30d; hosts
with no uptime yet last.
- Agent panel and fleet widget: last seen as "3m ago", not a clock time.
- SNMP: a device is "Reachable" only while it answered within the last 3
polls; older readings say "No answer" with their age, and every device
says when it was last read (was: reachable whenever any reading existed).
- Traefik: a router missing from 3 scrapes in a row is listed apart as no
longer reported, by when it was last seen, instead of keeping its last
state (a removed router stayed Failing).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5210 "Problems first, with since-time, across every list" (milestone 477),
first batch.
- Alerts: the Active list reads firing, no data, acknowledged, pending, the
longest-running first, and re-polls itself. An acknowledged alert says how
long it has been breaching (from the alert's events; acknowledging moves
last_transition_at) and, apart from that, who acknowledged it and when.
Rules show what each is doing now; disabled rules sort last. The current
value reads the latest-value table (#5208) instead of plugin_metrics.
- Monitors: a down monitor says "Down 12m · connection refused" in the list.
- Hosts: sorted down, agent gone quiet, pending, healthy. The agent column
shows "—" for a missing reading (was 0%) and a Stale badge with its age.
agent_metrics_by_host is gone; the list reads agent_overview_by_host.
- Schedules: failing first (failed or interrupted run, or a firing that could
not start one), then running, then the rest, paused last; status badge
from the shared states.
- Maintenance windows: active (ending soonest), then upcoming (starting
soonest), with expired folded away; each says when it ends/starts/ended.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
failing_services filtered on `(service_trouble(s) or 9) <= 2`; a crash loop
ranks 0, which is falsy, so the worst kind of trouble was left out. Test the
rank against None. The page test's task error also gains the age_s the
_swarm_failures include reads.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#5209 "Docker views: readable names, CPU/memory in cores and bytes, failures
listed with since-time" (milestone 477 "UI problems first").
- A swarm task reads as its service and slot ("codex_db · 1", or the node
for a global service), with the raw name in the title: Docker page, host
panel, busiest widget, container detail.
- CPU reads in cores (Docker's 100% = one core), toned by its share of the
host's CPUs from the agent's per-core series. Memory reads in bytes, against
a limit only when one is really set (Docker reports host RAM otherwise).
- Exit 137/143 without an OOM kill is a stop, not a failure: no
exited_nonzero, no crash-loop death, no "failed (24h)", dim exit code.
- "Failed (24h)" is a list (Docker page #failures, linked from the widget
and the tile): plain containers by their deaths, swarm services by Swarm's
failed tasks; a second manager's copy of a task death counts once.
- Failing swarm services lead Needs attention; services sort problems first
with since-times on the page, the widget and the Swarm page; nodes that
are down or drained sort first.
- Busiest widget orders by a 5-minute CPU average so the list holds still.
- Host panel: a table, problems first with what is wrong and since when;
replaced swarm tasks hidden and counted.
- Container detail: node hostname, I/O labelled "since start", Restarts
hidden for swarm tasks with a note on where task failures are.
- Widget names: "Docker — Services & failures", "Docker — Busiest containers".
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Migration 0030 seeded plugin_metric_latest from the whole history: two index
probes per distinct metric key. On the live instance that held the first
boot for 7+ minutes with Steward down. The migration now only creates the
table. Live sources fill it on their next sample, and backfill_latest runs
after startup for resources that have history but no latest rows. Per
resource it does one backward probe for the last sample, then takes the
newest value per metric within the hour before it. Resources that already
have rows are skipped, so later boots only walk the resource names.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
_last_n_results (Status, the monitors list, host monitors), host_status
(hosts overview, list_hosts) and the report's down-monitor list each
ranked every result a monitor ever recorded to keep the newest few. Each
monitor now gets ORDER BY checked_at DESC LIMIT n in a LATERAL subquery,
which walks ix_monitor_results_monitor_checked backwards and stops.
Also fixes the latest-table test, which read between two begin() blocks
and so autobegan a transaction the second block then refused.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The dashboard's host widgets took 18-60 s. Every "latest reading" query
(hosts overview, host-agent fleet, host page, vitals, SNMP, MCP get_host)
searched plugin_metrics for the newest row per metric. Postgres has no skip
scan, so each one read every raw sample in the 7-day window.
- plugin_metric_latest: one row per (source, resource, metric).
record_metrics upserts it in the same transaction, moving forward only and
in sorted key order. Migration 0030 seeds it by stepping the existing index.
- Retention drops a latest row once its sample leaves the raw window, so a
resource that stopped reporting disappears when its history does, as before.
Renames carry it along.
- A quiet host keeps its last values, marked stale. A time-bounded query
would have shown it as missing.
- Vitals sparklines span the 6 h they claim, averaged in SQL to ~120 points,
instead of fetching 6 h raw and drawing the last 120 samples.
- Dashboard and share-view cells: "Loading…" until the first answer; a 20 s
deadline per fetch; "Couldn't load (why)" with Retry now and an automatic
retry for cells that don't poll; "as of HH:MM" in the header, and
"couldn't refresh · as of" when a refresh fails over shown data.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The operator's concept, redrawn as a single-colour SVG in the accent gold:
a key whose bow is a cloverleaf holding three connected nodes, hanging on a
loop threaded through a keyring and swinging into the ring's middle, with a
small sparkle.
- static/logo/steward-logo.svg: the full mark (node graph in the bow).
- static/logo/steward-mark.svg: the small mark for 16-32 px (heavier lines,
one centre node, two teeth); the node graph turns to mush below ~40 px.
Used in the sidebar brand (was an inline crown) and as the favicon on every
page and on the public share view.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Run 8447 on b307f53: the boot check failed at the forced assertion, Push was
skipped, clean-up ran, :dev stayed on sha256:718d2ca (98854d9) and no
:b307f53 image exists in the registry.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Dockerfile: python:3.14-slim, the Python CI tests on (was 3.13). tini from
apt is PID 1 (ENTRYPOINT tini -- entrypoint.sh): it reaps the zombies ssh
and ansible leave and forwards SIGTERM.
- cli: Hypercorn's serve() with one worker (run, stream and scheduler state is
in-process), a 10 s graceful timeout, and SIGTERM/SIGINT as the shutdown
trigger. Access and error lines go through Steward's logging. --debug keeps
Quart's reloading dev server.
- Logging: steward/core/logging_setup.py installs one root handler before
anything logs (until now a library's basicConfig on import was the only
configuration). Its level is the `log.level` setting (DEBUG/INFO/WARNING/
ERROR, default INFO), applied at boot and on save in Settings → General.
- CI publish: build, then "The image boots with no internet" (rule 164,
snippet #3552), then push, so a broken image never reaches a tag (rule 177).
On an --internal network with its own Postgres, the image must migrate and
answer /health, run tini as PID 1, serve first-run setup with its assets
from /static and no third-party host, and exit 0 on SIGTERM. Logs and
clean-up run always.
Tests: tests/core/test_logging_setup.py; the smoke step itself.
Not in this commit: the dependency lockfile (it conflicts with rule 154 and
is waiting on the operator).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The "[queued — waiting for run X on the same host]" and run-slot lines were
broadcast to live viewers and never stored, so a run that sat queued had no
record of why once the live view was gone (CI 8439: the overlap test found
the order right and the explanation missing). _claim_hosts now returns the
line it waited on and start_run seeds the run's output with every wait.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- One run per host: start_run claims every host it touches (addresses read
from the inventory Steward wrote; a source's own inventory file is claimed
as that file) before it takes a run slot. The claim is all-or-nothing. A run
waiting on a host shows queued and names the run it waits for, honours a
cancel, and fails after HOST_WAIT_SECONDS (1 h) with which host and run.
- A due schedule whose previous run is still queued or running skips the
firing: last_run_at moves on, last_run_id still names the run going, and
last_error says "Skipped: the previous run was still going". The Schedules
page shows that beside the run (muted), and a launch error in red.
- An alert rule that fires while the action run from an earlier firing is
still going starts no new run; the event says so and names the run. An
in-process set closes the gap between the check and the run being linked.
- docs/core/ansible.md: queued/cancelled statuses and the overlap rules.
Tests: tests/core/test_ansible_overlap.py (claim keys, waiting, all-or-nothing,
cancel, deadline) and tests/integration/test_ansible_overlap.py (two real
runs on one host take turns; schedule skip then fire; alert action skip then
start).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- git ships in the image; checkouts live under /data/ansible/sources
(STEWARD_ANSIBLE_SOURCE_DIR), on the persistent volume.
- One task, ansible_git_pulls, reads the sources every 30 s and pulls each
as its interval passes: a source added in Settings is pulled without a
restart, a removed one is forgotten.
- git_pull: a 120 s timeout ending the process group, GIT_TERMINAL_PROMPT=0,
pull --ff-only, a failed clone removed; it raises GitPullError with git's
message, so the Sync button stops saying Synced after a failure (and the
message is now escaped).
- An alert action's vars and the firing's steward_alert_* vars reach
ansible-playbook as one JSON extra-vars file, not -e key=value, which
Ansible split on whitespace (a rule named 'High CPU on web').
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Output is read in 64 KiB chunks and split into lines here; a line past
64 KiB is cut and marked instead of raising out of readline's limit.
- ansible-playbook runs in its own session; whatever ends the read (finish,
an error, the task being cancelled) SIGTERMs then SIGKILLs its whole group,
so no ssh or worker it forked outlives the run.
- A run whose task is cancelled (shutdown) records 'cancelled' with
finished_at before re-raising, instead of staying 'running'.
- A cancel that lands after the process exited 0 no longer turns a
successful run into 'cancelled'.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- prune_by_age(column) registers a time-series column with the hourly cleanup;
Traefik (metrics, access summaries, global stats) and UniFi (WAN, client,
WLAN, DPI, events, alarms, speedtests) register theirs in setup().
- Alert events join monitor results and Ansible runs on data.retention_days,
read from Settings each run; Ansible run log files older than it are deleted.
- audit.retention_days (365) keeps the audit log longer, set on Settings > General.
- metric_series.points reads metrics.retention.raw_days, so the chart's
raw/rollup seam is where the rollup put it (RAW_DAYS removed).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
MAX_SAMPLES_PER_INGEST was 2 while agent 1.10 drains a backlog five samples a
post, so the server dropped the oldest three of every such post. Posts now
answer in 0.4-1.3 s on dev, including the boot cleanup window, so five fit
the agent's 10 s. A test holds the server cap at or above the agent's.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Agent 1.11.0 sends finished_at (State.FinishedAt) and reads a stopped
container's logs once per stop: since its cursor if it was being read, or
its last 200 lines when first seen already crashed (non-zero exit, stopped
in the last 10 min). Cursors and finals are forgotten once a container is no
longer listed.
Ingest gives a container first seen already exited non-zero a die event, and
dates die/stop events by finished_at when the agent sends it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
persist_host_docker marks docker.status, .swarm, .disk, .logs, .metrics,
.events, .containers, .alerts and .reap on the ingest's phase timer, each
flushed before its mark so a part's writes are charged to it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
An audit of 1,980 text/surface pairs (every steward.css rule and every inline
template style, tints composited over page, card and hovered row) found:
- Red, yellow and orange status text on its own 20% tint over a hovered row
at 3.97-4.06:1 (badges, count badge, alert-error, btn-warn). --red and
--yellow are lifted 15% toward the primary text, --orange 18%, as --blue
already was: worst case now 4.54:1.
- Inputs, selects, ghost and range buttons outlined in --border-mid at
1.4-1.9:1, on surfaces 1.1:1 apart, so the outline is what marks them.
New --border-control (45% toward secondary text): 3.13:1 worst. The
dashboard switcher and widget-drawer fields drop their inline styles for
the shared .input class.
- The breadcrumb separator painted in a border colour (1.4-1.9:1): now
--text-dim (4.64:1) and aria-hidden, since it is decoration.
tests/test_contrast.py resolves the tokens (color-mix and tints included)
and holds every drawn pair to 4.5:1, status text on its tint to 4.5:1 and
control outlines to 3:1, with a test that the guard catches the old pairs.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Measured on dev (385a579), agent posts cut to five samples still dropped:
flush=8.8s for 990 rows, 6.6s for 680, 8.4s for 1000 — about 110 rows/s,
~9 ms a row, with rules and evaluation near zero. One random read per
row fits a random uuid4 primary key whose index no longer fits in memory:
every new row lands on a random page of it. (Inferred from the rate; the
table and index sizes have not been read.)
- steward.core.ids.time_ordered_id: a UUIDv7-layout string (48-bit ms
timestamp, then random), same 36-char shape as the stored uuid4 ids, so
no migration. Python's uuid7() arrives in 3.14; the image runs 3.13.
- Used for plugin_metrics and its hourly rollup, monitor_results,
audit_events, the docker metrics/hourly/events/logs tables, and the
traefik and unifi time series.
- MAX_SAMPLES_PER_INGEST 5 → 2 (~400 rows) for now: it fits the agent's
10 s even at the old rate, so hosts land data whichever way the new rate
comes out. Raise it once the new rate is measured.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- BACKOFF_CAP 300 → 60: Steward reads a host stale after 180 s, so an agent
backing off to five minutes read stopped long after Steward came back
(lesson #4355: a check-in inherits its call's backoff).
- A POST carries at most five samples: the oldest four buffered plus the
new one, the rest held for the next post. A backlog drains a few at a
time instead of in one post Steward cannot write before the timeout
(the loop measured in #5149).
- Per-container Docker work has a 20 s deadline over the whole fan-out.
Past it the sample reports docker_status "error" ("N of M containers
not read within 20 s") and no list: a partial list would read as
removals, and a record without its inspect as restarts.
- /system/df is asked every 600 s rather than every sample; Steward keeps
the last answer.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Measured on dev (f802c9f): every dropped ingest carried samples=21,
metric_rows≈4,200 and never finished the INSERT flush inside the agent's
10 s ("completed: parse=0.1s"). The rollback left the agent's ring buffer
full, so each later post was the same size and failed the same way: no
host's metrics have landed since.
An ingest now keeps the newest MAX_SAMPLES_PER_INGEST (5) samples by ts,
logs how many it left out, and returns samples_dropped. The older samples
are the ones the ring buffer would have overwritten anyway. Why the
database writes so few rows a second is not yet known; the slow-ingest
line now measures it on a post that finishes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
