REPORT: v0.338.0 released (operator yes 2026-10-22): 3 test boxes through the floor, checks green, global floor set
gates / gates (push) Successful in 1m41s
gates / gates (push) Successful in 1m41s
This commit is contained in:
@@ -1,22 +1,23 @@
|
||||
# REPORT — controller, the night of 2026-10-21: D22 (the put-back for every restore), the first-hour words, R-1199 — COMMITTED, NOT RELEASED
|
||||
# REPORT — controller v0.338.0 released (2026-10-22 morning, operator yes)
|
||||
|
||||
**Brief:** the unattended night brief of 2026-10-21 (Parts 2–5). Architecture read: felhom.eu `07-backup-architecture.md`
|
||||
§6.3/§8 (the restore failure ladder, R-893 option A as built), `08-alarm-ladder.md` (notification groups),
|
||||
`10-localisation.md`. Test image **`0.338.0-rc.2`** (`sha256:ae52c93ebb83af1c646986119a1b74b07f992ffb03017624d052ece6266ec748`)
|
||||
from `f0abdd0`, smoke-tested on 9202 (every page 200, 0 error/panic lines), 9202 put back on 0.337.0. `0.338.0-rc.1`
|
||||
(`d582feb`) carried the D22 proof. **The release (0.338.0, MinAgent 0.158.1) is the operator's morning yes.**
|
||||
**What:** D22 — every restore puts the household's data back exactly on failure and after a power cut (R-1177, R-1195,
|
||||
R-1178, R-1180); the rehearsal's first-hour words (W9–W14, W21, W22, W24, W25, W28, W29); the install page's host-swap
|
||||
sentence; R-1199 (a box with no saved notification choice tells the hub the list its page shows — decided by CC the night
|
||||
before, operator may reverse). Proven the night of 2026-10-21 on 9202 (felhom.eu `REPORT-night-2026-10-21.md` §4).
|
||||
**MinAgent 0.158.1** (unchanged; the swap sentence stays silent with an agent older than 0.162.0). No golden (`09` §3
|
||||
decision 250).
|
||||
|
||||
| Row | What | Proof |
|
||||
|---|---|---|
|
||||
| R-1177 / D22 | The local unit restore measures, copies and puts back exactly on any failure (reuses `offbox_restore_undo.go`) | 9202 rc.1: damaged unit → live data back (A,C), app running; `TestR1177_*` red-proved |
|
||||
| R-1195 / D22 | A restore cut by a power loss / killed controller is put back at start (`restore_crash_putback.go`, marker `undo`) | 9202 rc.1: kill -9 after the volume step → A,C,D kept, app running; `TestR1195_*` (local + off-site) red-proved |
|
||||
| R-1178 | Second-drive whole restore asks the unit's refusals first | `TestR1178_*` red-proved |
|
||||
| R-1180 | No second-drive mirror of an app while a restore of it runs | `TestR1180_*` (production rsync) red-proved |
|
||||
| R-1199 | A box with no saved notification choice pushes the default list at start | `TestR1199_*` red-proved; *decided by CC unattended* |
|
||||
| R-1192 follow-up | Install page: „A gépen nincs elég cserehely (swap)…" when memory is tight and the host's swap is short | `TestR1192_SwapShort*` red-proved |
|
||||
| W9–W29 | First-hour words (family names with accents, drive name pre-filled, sharing asks for its password, off-site failure in Hungarian, operator detail behind „Haladó"/„Részletek") | per-row tests in CHANGELOG, red-proved |
|
||||
| W22 | Two restore refusals in Hungarian (`restore_busy`, `restore_no_drive`) | gates + parity |
|
||||
| — | Fixed: `TestR603_*` was red on main since 0.337.0 (two English apostrophes) | `go test ./...` green |
|
||||
|
||||
Commits: `d582feb`, `37b79ca`, `8c1e290`, `28b05d6`, `5776d54`, `ba22f9b`, `f0abdd0`. CI: green by job id for the
|
||||
pushes checked (2175, 2178; the rest in the night report). Evidence: felhom.eu `documentation/audits/night-2026-10-21/`.
|
||||
- **Built** `./build.sh 0.338.0 --push` from `7900a32` (clean, pushed): image digest
|
||||
`sha256:e8a529f395bfe8c0e7b87d5d98929a6f8e9834aa13c0ab77bd214ee9e7e53433`.
|
||||
- **The three test boxes through the normal update path:** a per-customer floor 0.338.0 (MinAgent 0.158.1) for Tester 1,
|
||||
demo-hp and demo-felhom → each box swapped its controller through its agent in about 10 s (`controller-swap: new
|
||||
controller healthy`, 07:54–07:55 UTC); the hub logged `controller_updated 0.337.0 → 0.338.0` for all three.
|
||||
- **The box walk's checks green on all three:** `https_only_check.py` — 28 public names, every one 301 to https, the
|
||||
negative control convicted; 10 / 24 / 9 containers, 0 unhealthy, 0 restarting; 0 `level=ERROR` lines in the new
|
||||
controller's log. Evidence felhom.eu `documentation/audits/morning-2026-10-22/release/`.
|
||||
- **Global floor** 0.338.0, MinAgent 0.158.1, with CC's hub token: `303 → /configuration?flash=floor_set`; read back
|
||||
from the hub DB with its WAL (`min_controller_version = 0.338.0`, declared `0.338.0=0.158.1`, 08:05:09 UTC). The three
|
||||
per-customer floors cleared afterwards (303 each).
|
||||
- **Found:** the install page's swap sentence compares the host's swap with the guest's RAM; on a guest larger than the
|
||||
installer's 8 GB host swap (demo-felhom 12 GB, demo-hp 25 GB) it will show when memory is tight although the measured
|
||||
burst used 2.7 GB. Same root as the agent/hub check — felhom.eu R-1216, an operator choice.
|
||||
|
||||
Reference in New Issue
Block a user