felhom-controller/REPORT.md

REPORT — v0.349.0 released as the global floor: the busy file and the household line for „Reboot guest" (D65 = A); no alarm after a hold clear (R-1350) (2026-10-29 evening, attended)

Plain words first. The controller 0.349.0 is the floor for every box since 19:24 CET. While it runs a backup, a restore, an off-site run, the night catch-up, an app update or an app export/import, it keeps a small „busy" file fresh, and the agent refuses the operator's „Reboot guest" until it ends. After an operator reboot the dashboard says „A Felhom ügyfélszolgálata HH:MM-kor újraindította a dobozt." for a day. When the operator clears a hold and the app stays stopped, no „Telepített alkalmazás nem fut" follows any more (R-1350).

Commits

Commit	CI (job id)	What
b69e90b	2521 green	internal/busymark (the busy file, 15 s, removed when idle and at start) wired with five probes; agentapi.OperatorReboot + web.WatchOperatorReboot + the dashboard line (HU/EN, dashboard.operator_reboot; the idle dashboard byte-identical); R-1350: stacks.Manager.RecordStopWantedIfDown run through AppStopGuard.ClearHoldByOperatorThen once the hold is found and before it is lifted. Tests red first (R-1350: build failure, then the mutants); 17 mutants killed (two first-run mutants did not compile and were re-made).
94049a7	2526 green	v0.349.0 CHANGELOG heading — the release build's commit. MinAgent 0.158.1.

Release

• rc.1 (b69e90b, sha256:3bd08409…) on 9202 (demo-hp): walk green (6 containers, 0 errors, login and file manager 200); on Tester 1 for the live proofs.
• 0.349.0 from the clean pushed tree at 94049a7: sha256:c5341883cd58410ea631fc0c76262d0902a731a24c1096df551e7d35df1ea877.
• Per-box floors 19:21 CET (Tester 1, demo-hp, demo-felhom; MinAgent 0.158.1): all three on 0.349.0 within 45 s; the walk green on all three (0 unhealthy, 0 refusals, 0 ERROR lines, login and file manager 200, the socket 0600) and https_only_check.py green.
• Global floor 19:24:23 CET on the operator's yes (303 floor_set; the hub log and the configuration page read 0.349.0 / 0.158.1).

Live proofs (Tester 1, rc.1)

• The busy file: an off-site run pressed from the hub → busy 11 s later („a backup, a restore or an off-site run is running"); the run ended 18:04:38 UTC, the file was gone at 18:04:47; the agent refused the reboot meanwhile.
• The household line: after the operator reboot the dashboard (logged in, box_walk) read „A Felhom ügyfélszolgálata 19:05-kor újraindította a dobozt." (ASCII control jraind present, and the page title).
• R-1350: a throwaway Gokapi in a crash loop → held 18:09:57 UTC → cleared from the hub → „desired state for gokapi recorded as "stopped" (was "running")" → still gone after the next settings save → no app_start_failed in 6.5 minutes (before the fix: 30 s after the clear). Positive control: the dead-app scan at 18:17:27 UTC read „20 scans since boot, 4 deployed app(s) evaluated, 0 currently down". Gokapi then removed through the product (no container, no app.yaml, no volume).

Evidence: felhom.eu documentation/audits/release-2026-10-29c/ (p1-reboot/, p3-r1350/, release-controller-0.349.0/).