system disk usage with details #40

2026-09-09T14:12:42+02:00

mi commented

2026-09-09 14:12:42 +02:00

No description provided.

mi added 7 commits 2026-09-09 14:12:42 +02:00

system: colour the chart lines by level in every range 6cfe6ba659

The value line of a gauge chart is cut at the 80 % and 90 % thresholds now
and every piece drawn in its own colour, the crossing interpolated, so a
stretch above watch is yellow and one above critical red — in the 1 h and
4 h ranges too, which had no coloured line at all. The maxima line of the
aggregated ranges follows the same rule instead of being permanently red:
its colour says something now.

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

system: show the range average next to a gauge's current value c576894579

The CPU and RAM cards say what the machine does right now; what it usually
does over the shown range was only readable from the chart. GET
/api/system/history carries the mean of every series now, and the card puts
it right-aligned beside the big current value, smaller and marked with the
average sign. Disks are left out on purpose: a disk's mean over an hour is
its current fill and says nothing.

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

system: give the sizing advice from one day on, marked preliminary e33aef4b2a

Waiting a full week before saying anything was too cautious: a day already
covers a whole daily build cycle, and a rough first answer beats none. The
advice is computed from one day of history now, and the basis line says
'preliminary, reliable from 7 d' until the week is full — the numbers
themselves are unchanged.

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

docs(prs): the system-page chart, average and sizing changes a4f4b1b6a9

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

docs(rfcs): RFC 0002 — disk usage breakdown and physical headroom f771b589e2

Evaluates the two System-page questions that were asked with 'this has to
be evaluated first': what fills a monitored disk, and how to show a volume
running out while the quota still has room. Includes the drafts.

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

docs(rfcs): record the RFC 0002 decisions — A1 and B1 ad10694a14

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

system: show what fills a disk, and mark the physical headroom 439fb802c4

Implements RFC 0002 as decided: A1, the breakdown as a dialog, and B1, the
volume marked inside the quota's gauge.

The throttled walk that produced the repository size now tallies each file
into the subtree it lies in — worktrees, build environments, the checkout —
and the artifact store is walked alongside, so the breakdown costs no second
walk. The slices are keyed by file store, because repositories and artifacts
may live on different filesystems; whatever the budget holds beyond them is
one 'other' slice, named as not Werkator's.

The volume stops competing with a quota for the same number. Where a quota
applies it is the budget and keeps the scale; the volume becomes a constraint
on it, marked on the bar, named in the note, and judged in the level. Without
a quota nothing changes.

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

mi added 1 commit 2026-09-09 14:16:10 +02:00

docs(prs): PR#40 — one document for the whole System-page change f553c332be

The chart, average and sizing changes and the disk-usage work went into one
pull request, so they get one PR-doc rather than two, and the placeholders
in the code and the scenario ids become 40.

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

mi merged commit 02896c7c96 into main

2026-09-09 14:18:16 +02:00

mi deleted branch system-disk-usage

2026-09-09 14:18:16 +02:00

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: mi/werkator#40