A system page that shows the machine's health at a glance #27

2026-09-08T10:19:57+02:00

mi commented

2026-09-08 10:19:57 +02:00

Co-Authored-By: Claude Fable 5.1 noreply@anthropic.com

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

mi added 1 commit 2026-09-08 10:19:57 +02:00

docs(prs): open the system-health page PR-doc 471c407ea2

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

mi added 1 commit 2026-09-08 10:20:06 +02:00

docs(prs): PR#27 has its number 01556e5d10

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

mi added 1 commit 2026-09-08 10:25:55 +02:00

docs(prs): PR#27 scenarios for the system health page c754efa910

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

mi added 5 commits 2026-09-08 10:43:04 +02:00

metrics: sample load average, swap and pressure-stall information 1c5e00e98d

The three kernel signals that tell an overloaded machine from a busy
one (Scenario#27.01); each is null where its source is unavailable.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

metrics: pronounce a machine verdict with reasons 25949bb2cc

SystemHealth.assess combines RAM and disk headroom, load per core,
pressure-stall time and swap into healthy/pressure/overloaded
(Scenario#27.02); CPU utilization alone never counts.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

metrics: keep a day of minutes and a month of quarter hours ee3042d764

MetricsHistory is a column-wise ring buffer persisted with the
aggregation state (Scenario#27.03): 1440 minute points, 2880 quarter
points with average and maximum, a restart continues both.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

server: serve the sample history with its peaks and the running builds b57a7a094c

GET /api/system/history?range=day|month (Scenario#27.04) and the
runningBuilds list on GET /api/system (Scenario#27.05).

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

server: the system page states the machine's health at a glance a5a8d498de

A verdict line, one gauge card per resource with a sparkline and the
peak of the selected range, a tile strip, and a 24 h / 30 d toggle
(Scenario#27.05); the all-time min/max/avg table is gone from the page.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

mi added 1 commit 2026-09-08 10:59:01 +02:00

server: revalidate static resources on every load d4e064c341

The first rollout rendered the new system page with the browser's
cached old stylesheet and script; no-cache, must-revalidate makes a
deploy visible at once at the cost of a 304 per load.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

mi added 1 commit 2026-09-08 11:14:17 +02:00

server: five history ranges and readable charts ab64dc985d

1 h, 4 h, 24 h, 7 d and 30 d; the charts are taller and carry a time
axis, so an hour of data on a day's axis is recognizable as such.
The PR-doc's Solution section, lost to a silent edit, is restored.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

mi added 2 commits 2026-09-08 11:25:02 +02:00

server: the tiles name what they show above the value 0255a0de53

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

metrics: sample every 15 seconds, an hour of samples by default 25fd8fa68f

Many builds last less than a minute and fell between two samples. The
history gains a third tier (15 s samples for four hours; minutes and
quarters carry average and maximum), the state file is written once a
minute and on shutdown, and the page opens on the last hour.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

mi marked the pull request as ready for review 2026-09-08 11:41:46 +02:00

mi closed this pull request

2026-09-08 11:41:58 +02:00

Pull request closed

Please reopen this pull request to perform a merge.

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: mi/werkator#27