Update arc: the undo and ladder spiked, the build plan for the 2026-09-23 rulings
gates / gates (push) Successful in 26s

- Part 1 (09 §6.1a, audit): the undo performed by hand on 9202 for docmost
  (PostgreSQL), romm (MariaDB) and vikunja (SQLite volume) - all three came
  back with data written before AND after the backup. The product's loader
  cannot do it: over a migrated PG database it fails on the new tables'
  foreign keys; over MariaDB it leaves them behind. A truncated PG copy loads
  rc 0 into an empty database. No-DB apps have no last-second copy.
- Part 2: one press jumps A -> C; the box's catalog clone is depth 1.
  Ladder format recommended: update_ladder in .felhom.yml, not git history.
- Part 3: memory watch red-proof results (harness change in the catalog repo).
- Part 4 (09 §6.4): ten parts, ~22 evenings; one open point (R-643).
- Rows R-637..R-644 opened; R-446/450/451/462/463 updated. STATUS, CONTEXT.

No product code. Live catalog untouched; 9202 back on it.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0159rPz1ZhFKsS53msqPYxtS
This commit is contained in:
adminandClaude Opus 5.5 committed 2026-09-23 08:54:41 +02:00
1 parent 805ad1e962
commit 4c92beab8f
67 files changed
+5977 -162

No files matched your search

+11 -3
View File
@@ -666,14 +666,14 @@ class (an image `VOLUME` at an unmounted path) is still live — `immich-server`
| **R-440** | **[P2-MEDIUM] 23 catalog image pins float, so an update is not reproducible.** `compose pull` on a moving tag fetches whatever upstream published that day. **MEASURED 2026-09-01 over `app-catalog-felhom.eu` @ `29edad9c5bf4`: 79 `image:` lines across 53 apps, 66 distinct; 23 of those lines carry a tag with no patch version.** `postgres:16-alpine` (8 apps), `redis:7-alpine` (6), `mariadb:11.6` (2), plus one each of `postgres:15-alpine`, `postgis/postgis:16-3.5-alpine`, `mariadb:11.4`, `mariadb:12.3`, `ghcr.io/claperco/claper:2.5`, `ghcr.io/thomiceli/opengist:1.13`, `wger/server:2.6`. **A 24th is arguable and is recorded rather than rounded away:** `ghcr.io/immich-app/postgres:16-vectorchord0.4.3-pgvectors0.2.0` pins both extensions exactly but leaves the PostgreSQL patch floating. A customer pressing Frissites can therefore swap their DATABASE ENGINE build with no catalog change and no record; two boxes updated on two days end up different. **Severity MEDIUM on its own; it becomes BLOCKING the moment a pre-update copy exists**, because "what did we upgrade from and to" must be recordable and today it is not — which is also why R-440 must be read next to the digest discipline in Rule 10 of the spike. **MEASURED LIVE 2026-09-01 — the floating pins have ALREADY moved, with a passing control.** Running digests on demo-hp compared against what the registry serves for the same tag today: **`mariadb:11.4` MOVED** (`sha256:4f1d8d20...` -> `sha256:611a2fcc...`) and **`mariadb:12.3` MOVED** (`sha256:a02fe89c...` -> `sha256:dd9b303a...`), while `postgres:16-alpine`, `redis:7-alpine`, `mariadb:11.6` and `opengist:1.13` were SAME — **and both fully-pinned CONTROLS (`rommapp/romm:5.0.0`, `privatebin/pdo:2.0.5`) were SAME.** So on a box with ZERO visible drift by tag, pressing Frissites today silently swaps the DATABASE ENGINE build under `romm` and `bookstack`, with no catalog change and no record. **Compounding fact found while reading:** the recovery unit records `ImagePins` but the manifest comment says *"image NOT stored - re-pulled on restore"*, so a RESTORE of a floating-pinned app also re-pulls whatever is current — the same non-reproducibility on the recovery path. **HALF OF THE ANSWER SHIPPED 2026-09-02 (controller v0.233.0, slice 1): `app.yaml.installed_images` now records, per compose SERVICE, the reference AND the repo digest each container was actually created from — so "what did we upgrade FROM" is answerable on any box that has taken one lifecycle action since the upgrade.** What is still missing is the other half: comparing that digest against what the registry serves for the same tag TODAY, which needs a network call the render path deliberately does not make (see R-446). **The row therefore stays OPEN and its rank is unchanged** — recording a digest does not make a floating pin reproducible; it makes the drift measurable after the fact. `audits/SPIKE-app-update-2026-09-01.md` | **OPEN — rank P2-MEDIUM; owner: CC** |
| **R-444** | **[P3-LOW] Nothing runs `pct fstrim` on the fleet, and demo-hp's thin pool was carrying ~23.8 GB of blocks the guest had already freed.** MEASURED 2026-09-01 during this spike's teardown: the run itself added ~1.05 GiB that `local-lvm` did not reclaim on delete (68.97% -> 70.91%); `fstrim` INSIDE the unprivileged container is refused (`FITRIM ioctl failed: Operation not permitted`, all three mounts); `pct fstrim 9201` from the PVE host then trimmed **30.2 GiB + 57 GiB** and took `local-lvm` to **26.78%** — **23.8 GB BELOW this run's own starting point**, i.e. the surplus was long-standing, not ours. **Why it is not merely housekeeping:** a thin pool that only ever grows can reach 100% from DELETED data alone, and a full thin pool takes every guest on the host read-only. demo-hp had 16.4 GB free before the trim. **Not urgent, and the row says so** — but the appliance has no periodic trim and no operator surface reports the gap between guest-free and pool-used. Owner: **CC.** `audits/SPIKE-app-update-2026-09-01.md` | **OPEN — rank P3-LOW; owner: CC** |
| **R-445** | **[P3-LOW] Hub app telemetry survives the app's removal, so a 15-minute throwaway now sets a FLEET-WIDE memory recommendation.** MEASURED 2026-09-01: this spike's Phase 6 Nextcloud existed for ~15 minutes on demo-hp, spent part of it crash-looping, and was then removed with all volumes. The hub's `/apps/nextcloud` page still reports `Deployments`, `Avg Memory 208 MB`, `P95 Memory 280 MB` and **`Suggested Limit (P95x1.2) = 352 MB`**, plus three MariaDB `io_uring` rows under Known Issues attributed to demo-hp. **The suggested limit is an operator-facing recommendation derived from a sample that no longer exists anywhere** — and Nextcloud is a real catalog app whose limit someone may act on. **RETAINED DELIBERATELY BY THIS RUN, NOT CLEARED, and the reason is part of the row:** the hub offers `POST /apps/nextcloud/reset-telemetry` whose own confirm reads *"Delete all telemetry data for nextcloud? This cannot be undone."* — an irreversible write on the operator's surface, and the operator authorised Phase 6, not this. **The one-line command is recorded in the audit doc so it is a decision, not a task.** The general question is the row: should telemetry for an app with zero live deployments age out, or be excluded from the suggestion? Owner: **VIKTOR rules, CC implements.** `audits/SPIKE-app-update-2026-09-01.md` | **OPEN — rank P3-LOW; owner: VIKTOR rules, CC implements** |
| **R-446** | **[P2-MEDIUM] „Naprakész" can be FALSE, and the badge that says it cannot tell.** Slice 2 (controller v0.233.0, 2026-09-02) compares the RECORDED image reference per compose service against the reference the current template pins, and **queries no registry** — deliberately: a customer's box must not depend on reaching eight upstream registries to render a page (`felhom-controller/controller/internal/web/updatebadge.go`, `compareInstalledToTemplate`). **For the 23 floating pins that comparison is blind by construction:** `postgres:16-alpine`, `mariadb:11.6` and 21 others can carry an identical reference over an image that has moved. **MEASURED, not theorised — spike §5 found `mariadb:11.4` and `mariadb:12.3` had BOTH already moved upstream while two fully-pinned CONTROLS held.** So `romm` and `bookstack` on demo-hp would read „Naprakész" over a database engine build that is not the one the catalog now resolves to. **This is a KNOWN LIMITATION OF A SHIPPED FEATURE, filed the same session rather than left implicit**, and it is stated in the same words in `architecture/09-update-architecture.md` §8.1 and in the controller's `README.md`. The close is a digest comparison against the registry, which needs a network call, a cache and a failure posture — it is not a one-liner and it is not slice 2's job. **Depends on R-440**, whose fix (stop floating) would remove the problem instead of measuring it — take that route first if it is available. `architecture/09-update-architecture.md` **MEASURED 2026-09-21, and the blind spot is not one or two pins.** `audits/UPDATE-ARC-STATE-2026-09-21.md` §3.3: the catalog carries **10 floating pins of 66** (recounted — the old "23" was stale), and **6 of the 7 measurable engine pins have been repushed upstream since the catalog set them** — `postgres:16-alpine` (8 apps), `postgres:15-alpine`, `redis:7-alpine` (6 apps), `mariadb:11.4`, `mariadb:12.3`, `postgis:16-3.5-alpine`; only `mariadb:11.6` has not. The 8th (immich's own ghcr build) is UNMEASURED — ghcr exposes no anonymous last-modified timestamp. **So on demo-hp today four apps read „Naprakész" over a database engine image that has demonstrably moved.** The fix does NOT need the box to query a registry: the catalog can record each pin's digest at push time (`check-image-resolvable.py` already resolves it) and the box compares digests. Put to the operator as `09` §3b **Q6**, recommended YES — the cheapest real improvement on the arc's list. **— UPDATE NIGHT 2026-09-21:** **MEASURED ON A BOX 2026-09-21 (update night, leg B8), and it REFINES the row in two ways rather than merely confirming it.** §8.1's numbers came from a registry sweep on DooPlex; this is the same question asked of a customer-shaped box, where the badge actually renders. On guest 9202, `docmost`'s two floating pins were read as `installed_images` records them and compared against the upstream digests measured the same night: `postgres:16-alpine` → **`sha256:721873c34ceb9…` on the box and `sha256:721873c34ceb9…` upstream**, and `redis:7-alpine` → **`sha256:858f009f9709c…` both sides**. **Identical. So the badge „Naprakész" is TRUE for this box**, and the app reads correctly. **(1) The defect's size is set by INSTALL AGE, not by the catalog.** A floating pin is wrong only for a box that pulled BEFORE the tag moved; a box deployed after the repush holds the current image and its badge is right. R-446's "six repushed pins" measured the tag against the date the CATALOG set it, which is the right measure for *the catalog* and not for *a box*. **(2) The producer Q6 needs ALREADY EXISTS on the box.** `installed_images` records a real `digest` per service (`installed.go` §7.1) — the box knows exactly what it is running. What it cannot do is COMPARE, because the catalog carries no digest to compare against. That is Q6's proposal, and this is a concrete confirmation that only the catalog half is missing. Evidence: `audits/update-night-2026-09-21/23-B8-floating-pin.txt`. **-- RULED 2026-09-23 (`09` §3 decision 17):** YES — the catalog records the image digest of every pin at push time; the box compares against it and, where the catalog carries one, pulls **that exact image**, which makes a floating tag reproducible, not only the badge honest. *Pull-by-digest while the definition names a tag is a claim to verify in the build, not a ruling on mechanism.* | **READY TO BUILD — owner: CC; `09` §6.4** |
| **R-450** | **[P2-MEDIUM] UPDATE ARC SLICE 6 — a version sequence: automatic WITHIN a major, never ACROSS one, and an engine change gets its OWN edge.** The first half is an operator ruling of 2026-09-02 and its justification is R-449's measurement: a cross-major jump can be refused by the app itself and cannot be undone. **The second half is a rule recorded now, while it is cheap:** an engine change must never be bundled with an app version bump. `bookstack`'s `0b73e5e` moved the application 25.02.2 → 26.05.2 **and** MariaDB 11.6 → 12.3 in one commit — **two migrations behind one edge**, and an unreadable failure when it breaks. Needs a catalog-side convention and, eventually, a gate. `architecture/09-update-architecture.md` §6 **HALF SHIPPED 2026-09-21 (catalog `5ff36d098cbc`): the second half — an engine change gets its OWN edge — is now ENFORCED** by `check-engine-major.py`, which refuses a commit moving a MariaDB major together with any other image move in that template, naming what it was bundled with. The FIRST half (automatic within a major) is Slice 6 and needs four operator answers — `09` §3b **Q1–Q4**, with the shape it would take in `09` §6.2. **The urgency is now measured:** 46 of the catalog's 58 exact pins are behind upstream and **39 of those are within a major** — the population the 2026-09-02 ruling already says may move without a human. **-- RULED 2026-09-23 (operator, `09` §3 decisions 11–15):** Q1–Q4 answered. The update is a leg of the backup chain after off-site and before the full-system backup (11); automatic with a per-box switch ON by default (12); **the TEST decides, not the tag** — the box applies every step the catalog holds because the catalog holds only tested steps, and `CompareImageRefs` moves to the catalog gate (13, REPLACES "never across a major"); a box behind climbs **one tested step at a time** (14); **the box UNDOES a failed update itself** — old definition + the pre-pin safety dump + health check again, HOLD only if the undo fails (15, REPLACES §6.1's no-auto-undo). The undo and the ladder were SPIKED the same day before any build (`audits/update-rulings-2026-09-23/`); build order and costs in `09` §6.4. | **READY TO BUILD — owner: CC; `09` §6.4 part by part, each part returns to the operator for go/no-go** |
| **R-446** | **[P2-MEDIUM] „Naprakész" can be FALSE, and the badge that says it cannot tell.** Slice 2 (controller v0.233.0, 2026-09-02) compares the RECORDED image reference per compose service against the reference the current template pins, and **queries no registry** — deliberately: a customer's box must not depend on reaching eight upstream registries to render a page (`felhom-controller/controller/internal/web/updatebadge.go`, `compareInstalledToTemplate`). **For the 23 floating pins that comparison is blind by construction:** `postgres:16-alpine`, `mariadb:11.6` and 21 others can carry an identical reference over an image that has moved. **MEASURED, not theorised — spike §5 found `mariadb:11.4` and `mariadb:12.3` had BOTH already moved upstream while two fully-pinned CONTROLS held.** So `romm` and `bookstack` on demo-hp would read „Naprakész" over a database engine build that is not the one the catalog now resolves to. **This is a KNOWN LIMITATION OF A SHIPPED FEATURE, filed the same session rather than left implicit**, and it is stated in the same words in `architecture/09-update-architecture.md` §8.1 and in the controller's `README.md`. The close is a digest comparison against the registry, which needs a network call, a cache and a failure posture — it is not a one-liner and it is not slice 2's job. **Depends on R-440**, whose fix (stop floating) would remove the problem instead of measuring it — take that route first if it is available. `architecture/09-update-architecture.md` **MEASURED 2026-09-21, and the blind spot is not one or two pins.** `audits/UPDATE-ARC-STATE-2026-09-21.md` §3.3: the catalog carries **10 floating pins of 66** (recounted — the old "23" was stale), and **6 of the 7 measurable engine pins have been repushed upstream since the catalog set them** — `postgres:16-alpine` (8 apps), `postgres:15-alpine`, `redis:7-alpine` (6 apps), `mariadb:11.4`, `mariadb:12.3`, `postgis:16-3.5-alpine`; only `mariadb:11.6` has not. The 8th (immich's own ghcr build) is UNMEASURED — ghcr exposes no anonymous last-modified timestamp. **So on demo-hp today four apps read „Naprakész" over a database engine image that has demonstrably moved.** The fix does NOT need the box to query a registry: the catalog can record each pin's digest at push time (`check-image-resolvable.py` already resolves it) and the box compares digests. Put to the operator as `09` §3b **Q6**, recommended YES — the cheapest real improvement on the arc's list. **— UPDATE NIGHT 2026-09-21:** **MEASURED ON A BOX 2026-09-21 (update night, leg B8), and it REFINES the row in two ways rather than merely confirming it.** §8.1's numbers came from a registry sweep on DooPlex; this is the same question asked of a customer-shaped box, where the badge actually renders. On guest 9202, `docmost`'s two floating pins were read as `installed_images` records them and compared against the upstream digests measured the same night: `postgres:16-alpine` → **`sha256:721873c34ceb9…` on the box and `sha256:721873c34ceb9…` upstream**, and `redis:7-alpine` → **`sha256:858f009f9709c…` both sides**. **Identical. So the badge „Naprakész" is TRUE for this box**, and the app reads correctly. **(1) The defect's size is set by INSTALL AGE, not by the catalog.** A floating pin is wrong only for a box that pulled BEFORE the tag moved; a box deployed after the repush holds the current image and its badge is right. R-446's "six repushed pins" measured the tag against the date the CATALOG set it, which is the right measure for *the catalog* and not for *a box*. **(2) The producer Q6 needs ALREADY EXISTS on the box.** `installed_images` records a real `digest` per service (`installed.go` §7.1) — the box knows exactly what it is running. What it cannot do is COMPARE, because the catalog carries no digest to compare against. That is Q6's proposal, and this is a concrete confirmation that only the catalog half is missing. Evidence: `audits/update-night-2026-09-21/23-B8-floating-pin.txt`. **-- RULED 2026-09-23 (`09` §3 decision 17):** YES — the catalog records the image digest of every pin at push time; the box compares against it and, where the catalog carries one, pulls **that exact image**, which makes a floating tag reproducible, not only the badge honest. *Pull-by-digest while the definition names a tag is a claim to verify in the build, not a ruling on mechanism.* **-- 2026-09-23:** pull-by-digest MEASURED on 9202 — Docker and Compose both pull and run `redis:7-alpine@sha256:858f…` and refuse a digest that does not exist. **Build trap, read from source:** `splitImageRef` returns "unorderable" for any ref containing `@` (`stacks/updateorder.go:134`), so a digest-carrying pin must have its digest split off before ordering or every such app reads Unknown. `09` §6.4 part 6. | **READY TO BUILD — owner: CC; `09` §6.4** |
| **R-450** | **[P2-MEDIUM] UPDATE ARC SLICE 6 — a version sequence: automatic WITHIN a major, never ACROSS one, and an engine change gets its OWN edge.** The first half is an operator ruling of 2026-09-02 and its justification is R-449's measurement: a cross-major jump can be refused by the app itself and cannot be undone. **The second half is a rule recorded now, while it is cheap:** an engine change must never be bundled with an app version bump. `bookstack`'s `0b73e5e` moved the application 25.02.2 → 26.05.2 **and** MariaDB 11.6 → 12.3 in one commit — **two migrations behind one edge**, and an unreadable failure when it breaks. Needs a catalog-side convention and, eventually, a gate. `architecture/09-update-architecture.md` §6 **HALF SHIPPED 2026-09-21 (catalog `5ff36d098cbc`): the second half — an engine change gets its OWN edge — is now ENFORCED** by `check-engine-major.py`, which refuses a commit moving a MariaDB major together with any other image move in that template, naming what it was bundled with. The FIRST half (automatic within a major) is Slice 6 and needs four operator answers — `09` §3b **Q1–Q4**, with the shape it would take in `09` §6.2. **The urgency is now measured:** 46 of the catalog's 58 exact pins are behind upstream and **39 of those are within a major** — the population the 2026-09-02 ruling already says may move without a human. **-- RULED 2026-09-23 (operator, `09` §3 decisions 11–15):** Q1–Q4 answered. The update is a leg of the backup chain after off-site and before the full-system backup (11); automatic with a per-box switch ON by default (12); **the TEST decides, not the tag** — the box applies every step the catalog holds because the catalog holds only tested steps, and `CompareImageRefs` moves to the catalog gate (13, REPLACES "never across a major"); a box behind climbs **one tested step at a time** (14); **the box UNDOES a failed update itself** — old definition + the pre-pin safety dump + health check again, HOLD only if the undo fails (15, REPLACES §6.1's no-auto-undo). The undo and the ladder were SPIKED the same day before any build (`audits/update-rulings-2026-09-23/`); build order and costs in `09` §6.4. **-- SPIKED 2026-09-23 (`audits/update-rulings-2026-09-23/`):** the undo works by hand on three real migrating edges and needs eight product additions (R-637..R-642); **the ladder is measured absent** — one press on a box two steps behind jumped vikunja 2.3.0 → 2.5.0 in 9.5 s and 2.4.0 never ran, and the box cannot see intermediate steps at all because its catalog clone is `--depth 1` (`sync.go:283`/`:300`, one commit visible on both demo guests). The ladder's recommended format is an `update_ladder:` list in `.felhom.yml` with each intermediate step's own definition, NOT the git history (romm's image-moving commit is the definition that OOM-looped). The chain's update leg has ≤15 min as ruled (R-643). Build order `09` §6.4. | **READY TO BUILD — owner: CC; `09` §6.4 part by part, each part returns to the operator for go/no-go** |
| **R-451** | **[P3-LOW] UPDATE ARC SLICE 7 — a fleet sweep: the operator can SEE, and MOVE, how far behind every box is.** Slices 1 and 2 make one box's state visible on that box's own pages. The operator has no fleet view, and **it is not derivable from what is already reported: the hub's report payload carries container name, state, CPU and memory, and NO image field at all** (spike §5, which is why Peti's box could only be recorded UNKNOWN). So this is a hub-side change as well as a controller one. Rank LOW today because the fleet is two enrolled boxes; it rises with the fleet. `architecture/09-update-architecture.md` §6, §8.4 **BOTH SIDES VERIFIED 2026-09-21, and it is cheaper than this row implies.** The controller's payload carries no image (`internal/report/types.go` L98–103) and the hub's `Store.SaveReport` (`hub/internal/store/store.go:965`) denormalises only container **counts** — but **the hub stores the raw report JSON whole**, so a new controller field lands there the day it is sent. What is missing is the denormalisation and the page, not the transport. Shape in `09` §6.2–6.3; the payload question is `09` §3b **Q7**. **-- RULED 2026-09-23 (`09` §3 decision 18):** the report carries, per compose service (database included), the installed reference, the catalog reference and the badge state. **Built later, when the fleet grows** — Q7's recommendation, confirmed. | **RULED — build deferred until the fleet grows; owner: CC** |
| **R-454** | **[P3-LOW] Five `internal/web` test files have been `gofmt`-unclean for an unknown length of time, and nothing notices.** MEASURED 2026-09-02: `gofmt -l controller/internal/web/` reports `backups_split_test.go`, `claim_code_naming_test.go`, `disk_health_test.go`, `r400_debug_routes_test.go`, `recovery_test.go` — at the **baseline** commit `960d29b0612c`, i.e. not introduced by v0.233.0 (both files added that day are clean). **`go vet` does not check formatting and `controller_gates.py` has no formatting gate**, so the only thing that would ever surface this is someone running `gofmt -l` by hand, which is how it was found. **Not reformatted in the same session, deliberately** — the minimal-changes rule, and a five-file whitespace commit inside a feature release makes that release's diff unreadable. **Small, and the cost of NOT having the instrument is the row:** the count can only grow, and every future `gofmt -l` run produces noise that hides a real one. Fix is two lines: a `gofmt -l` gate in `controller_gates.py` plus one formatting commit, in that order (the gate first, so the commit is provably complete). Owner: **CC.** | **READY — rank P3-LOW; owner: CC** |
| **R-457** | **[P3-LOW] A test that hardcodes a date AND asserts an age derived from it is green on the day it is written and red the next morning — one instance PROVEN, six candidate files named.** MEASURED 2026-09-03: `TestGroupD_BadgeRendersOnBothSurfaces` (shipped the previous day in v0.233.0) pinned a fixture `catalog_since: "2026-07-18"` and asserted the rendered string `"Frissítés elérhető — 46 napja"`. **The pure badge tests inject a clock; the RENDER test does not and cannot** — it goes through the production templates, which call the funcmap entry `updateBadge`, which reads `time.Now()`. The suite was green on 2026-09-02 and **FAILED on 2026-09-03** with *"the behind badge is missing"* on both surfaces, because the true answer had become 47. **Fixed by DERIVING the fixture** — `catalog_since` is computed as *today minus 46 days*, so the test asserts the real number through the real clock and cannot rot. **THE CLASS, which is why this is a row and not just a fix:** a clock-reading test that also carries a date LITERAL is a bomb with a fuse of unknown length, and the suite being green is not evidence it is defused — it is evidence the fuse has not burned down yet. **NAMED AS UNCHECKED CANDIDATES, NOT ACCUSED** — six other test files contain both a `20xx-xx-xx` literal and `time.Now()`: `internal/backup/offbox_test.go`, `internal/web/handler_export_upload_test.go`, `internal/web/r103_tier2_action_test.go`, `internal/web/dashboard_backup_card_test.go`, `internal/web/async_restore_test.go`, `internal/stacks/installed_test.go`. Mixing the two is not itself a defect — it is one only where a literal feeds an assertion evaluated against the real clock — so each needs reading, which is a sweep and not this session. **The instrument that would end the class:** run the suite once under a faked future date in CI and see what turns red. Owner: **CC.** `felhom-controller` v0.234.0 CHANGELOG | **READY — rank P3-LOW; owner: CC** |
| **R-458** | **[P3-LOW] `.felhom.yml` keeps flowing to an app whose compose file is FROZEN, so a frozen app can receive a health check written for a version it is not running.** The v0.235.0 render freezes `docker-compose.yml` for a pinned app once the catalog moves past its version, but copies `.felhom.yml` **verbatim in every case** (`Syncer.copyTemplates`). **The asymmetry is deliberate and both directions were considered:** `.felhom.yml` carries no image, and it carries `catalog_since` — the single input the update badge uses to say *„Frissítés elérhető — N napja"* — so freezing it would silently withhold the one number that tells a customer they are behind, i.e. it would break slice 2 to protect slice 3. **What it costs:** the file also carries the controller-side `healthcheck:` block and resource hints, so a template updated for a newer version can hand a frozen app a probe written for software it is not running. **THE FAILURE DIRECTION IS A FALSE ALARM, NEVER DATA LOSS** — the app keeps running; at worst it renders as degraded and, if it persisted, could reach the dead-app alarm path. That is the same class as R-330's false e-mails, which is why this is a row and not a footnote. **Not fixed now, and the reason is that the cheap fix is wrong:** freezing the whole file breaks the badge, and freezing only the `healthcheck:` key means the syncer would have to parse and re-assemble a customer-facing metadata file — new surface on the one path that touches every app on every box every 15 minutes. **What would settle it:** whether any catalog `healthcheck:` has ever been changed in the same commit as an `image:` line (measurable from the catalog's own history, no box needed). If the answer is "never", the exposure is theoretical and the row can be closed by measurement instead of by code. Owner: **CC.** `architecture/09-update-architecture.md` §5.4, §8.5 **— UPDATE NIGHT 2026-09-21:** **MEASURED 2026-09-21 (update night), leg B9, and the row's risk is NARROWER than it states.** A `.felhom.yml`-only change (a health check for a path only a newer version would serve) was pushed to a FROZEN `bentopdf` — installed `v2.8.6`, catalog ahead. §5.4's asymmetry is confirmed live: the new `.felhom.yml` reached the box while the compose `image:` line stayed `v2.8.6`. **But no false alarm was produced**: ten samples over two minutes all read `state=running` with the front door at `200`. The reason is the probe's own semantics, not luck — `healthprobe.go:258-261` treats **any response** as healthy for `type: http`, and the bogus path answers 404, which is a response. **So this row's false-alarm risk exists only for `type: api` probes carrying an `expect` block**, where the status is compared; for every `type: http` template and every `type: api` without `expect`, a newer version's path is invisible to the probe. The row's actual claim — the failure direction is a false alarm, never data loss — stands and is now measured. Evidence: `audits/update-night-2026-09-21/21-B9-frozen-app-newer-felhomyml.md`. | **READY — rank P3-LOW; owner: CC** |
| **R-460** | **[P3-LOW] BookStack's FILE half cannot be seeded or verified without a browser, so its upgrades can only ever be auto-proven for the DATABASE.** MEASURED 2026-09-06 while building the R-449 harness. BookStack's API needs a token that is only mintable through its web UI, and its HTTP login is unusable headlessly for a second, independent reason: `APP_URL` comes from the template as `https://${SUBDOMAIN}.${DOMAIN}`, so the app marks its session and XSRF cookies **`secure`**; curl over plain http stores neither and **every login POST returns 419 Page Expired**, which looks exactly like a wrong password. The container serves no TLS. **The database half IS provable** — the harness seeds with `php artisan bookstack:create-admin` and reads back with a DIFFERENT artisan command that must find the record, carrying its own negative control on every call. **What is unprovable is an uploaded image or attachment**, i.e. exactly the half a customer would notice. **THIS IS A FACT ABOUT THE APP, NOT A DEFECT IN THE HARNESS**, and it is recorded because Slice 6 needs to know which apps can be auto-verified and which can only be partly verified — nobody had that list before. **Deliberately NOT worked around:** planting a file in the volume would make the test pass while proving nothing, which is R-156's exact failure. **What would remove it:** a headless token route (upstream), or accepting a browser-driven step for this app alone, which DooPlex cannot run. Owner: **CC.** `audits/SPIKE-upgrade-test-2026-09-06.md` §6 **-- UPDATE NIGHT 2026-09-21:** bookstack's edge was walked again on 2026-09-21 and is again **half-proven**: the database half read back through `php artisan` with its own negative control, the file half untouched. The limitation is unchanged and is now measured on the box as well as on the harness. Two more apps joined the same class tonight for a different reason (R-624). | **READY — rank P3-LOW; owner: CC** |
| **R-462** | **[P2-MEDIUM] Widen the upgrade harness beyond three apps — and the cost is dominated by FIXTURES, not by machine time.** The R-449 harness works and is proven by a red negative control (`audits/SPIKE-upgrade-test-2026-09-06.md` §1). **Costed with this run's REAL numbers rather than an estimate:** a successful edge takes **6.4 s – 305.1 s, median 71.8 s**; a FAILING edge takes **556 s**, roughly **8×**, because a negative is only honest if it waits out the full settle window; 3 apps / 11 images cost **5.07 GB**, so 53 apps naively extrapolate to **~90 GB** and, at the median, about an hour of harness time for one edge each. **THAT EXTRAPOLATION UNDERSTATES THE REAL COST BY AN ORDER OF MAGNITUDE, and that is the point of this row.** Two of the three apps needed a bespoke non-browser seed route; one needed two attempts and a discarded approach; one (bookstack) can only ever be half-proven (R-460). **Fixture time scales with apps and does not amortise.** **The decision this row is really asking for is scope, not schedule:** all 53, or only the apps a customer would lose data from, or only apps whose catalog transition is a MAJOR. **Recommended shape, NOT a design — the operator picks:** start with the apps that carry a database, because §3 measured that the abort question only ever bites there. Owner: **VIKTOR rules on scope, CC implements.** `audits/SPIKE-upgrade-test-2026-09-06.md` §5 **ROW CORRECTED 2026-09-21: the scope is NOT open and this row said it was.** It read *"VIKTOR rules on scope"*; the operator ruled on 2026-09-13 (`09` §3 decision 6) that the upgrade test goes to **ALL** apps through the nightly rotation, explicitly *not* "database apps first". What is open is the WORK, not the scope. An ORDER inside that ruling — the 15 database services first, because that is where a wrong answer costs data rather than uptime — is costed as a drill brief in `09` §6.4: legs A–E ≈ **21–34 CC-hours** plus ~25–30 GB on a scratch host, with legs C (a PostgreSQL `pg_upgrade` rehearsal) and E (one automatic night on a throwaway) the two that unblock a decision. **-- UPDATE NIGHT 2026-09-21:** **The count moved from 3 apps to 21 EDGES ACROSS 19 APPS.** The update night walked real within-a-major upstream edges on scratch guest 9202 through the product's own guarded Update, each app seeded and read back through its OWN front door with a negative control on every readback: **14 proven, 3 failed, 4 inconclusive**. Proven: actualbudget, audiobookshelf, bookstack, docmost, grafana, home-assistant, mealie, n8n, navidrome, papra, privatebin, romm, vikunja, and nextcloud's MariaDB engine major. **Ten of the fourteen printed a verbatim migration line**, so the database really was rewritten and the data still read back. Box-side fixtures for 20 apps now exist at `audits/update-night-2026-09-21/fixtures.py`, and four (actualbudget, navidrome, audiobookshelf, vikunja) are ported into `app-catalog-felhom.eu/scripts/upgrade_fixtures.py` with seven new EDGES (U1-U7) so the same edges can be run on the harness venue **with their ABORT step**, which the box deliberately does not offer. **OWED, stated so it is not mistaken for done:** the U1-U7 harness RUNS (the code is in, the runs are not), and fixtures for the four inconclusive apps, of which two (vaultwarden, zipline) cannot be seeded at all while the catalog rightly closes their sign-up (see R-624). | **READY — rank P2-MEDIUM; owner: CC (scope already ruled, `09` §3 decision 6)** |
| **R-462** | **[P2-MEDIUM] Widen the upgrade harness beyond three apps — and the cost is dominated by FIXTURES, not by machine time.** The R-449 harness works and is proven by a red negative control (`audits/SPIKE-upgrade-test-2026-09-06.md` §1). **Costed with this run's REAL numbers rather than an estimate:** a successful edge takes **6.4 s – 305.1 s, median 71.8 s**; a FAILING edge takes **556 s**, roughly **8×**, because a negative is only honest if it waits out the full settle window; 3 apps / 11 images cost **5.07 GB**, so 53 apps naively extrapolate to **~90 GB** and, at the median, about an hour of harness time for one edge each. **THAT EXTRAPOLATION UNDERSTATES THE REAL COST BY AN ORDER OF MAGNITUDE, and that is the point of this row.** Two of the three apps needed a bespoke non-browser seed route; one needed two attempts and a discarded approach; one (bookstack) can only ever be half-proven (R-460). **Fixture time scales with apps and does not amortise.** **The decision this row is really asking for is scope, not schedule:** all 53, or only the apps a customer would lose data from, or only apps whose catalog transition is a MAJOR. **Recommended shape, NOT a design — the operator picks:** start with the apps that carry a database, because §3 measured that the abort question only ever bites there. Owner: **VIKTOR rules on scope, CC implements.** `audits/SPIKE-upgrade-test-2026-09-06.md` §5 **ROW CORRECTED 2026-09-21: the scope is NOT open and this row said it was.** It read *"VIKTOR rules on scope"*; the operator ruled on 2026-09-13 (`09` §3 decision 6) that the upgrade test goes to **ALL** apps through the nightly rotation, explicitly *not* "database apps first". What is open is the WORK, not the scope. An ORDER inside that ruling — the 15 database services first, because that is where a wrong answer costs data rather than uptime — is costed as a drill brief in `09` §6.4: legs A–E ≈ **21–34 CC-hours** plus ~25–30 GB on a scratch host, with legs C (a PostgreSQL `pg_upgrade` rehearsal) and E (one automatic night on a throwaway) the two that unblock a decision. **-- UPDATE NIGHT 2026-09-21:** **The count moved from 3 apps to 21 EDGES ACROSS 19 APPS.** The update night walked real within-a-major upstream edges on scratch guest 9202 through the product's own guarded Update, each app seeded and read back through its OWN front door with a negative control on every readback: **14 proven, 3 failed, 4 inconclusive**. Proven: actualbudget, audiobookshelf, bookstack, docmost, grafana, home-assistant, mealie, n8n, navidrome, papra, privatebin, romm, vikunja, and nextcloud's MariaDB engine major. **Ten of the fourteen printed a verbatim migration line**, so the database really was rewritten and the data still read back. Box-side fixtures for 20 apps now exist at `audits/update-night-2026-09-21/fixtures.py`, and four (actualbudget, navidrome, audiobookshelf, vikunja) are ported into `app-catalog-felhom.eu/scripts/upgrade_fixtures.py` with seven new EDGES (U1-U7) so the same edges can be run on the harness venue **with their ABORT step**, which the box deliberately does not offer. **OWED, stated so it is not mistaken for done:** the U1-U7 harness RUNS (the code is in, the runs are not), and fixtures for the four inconclusive apps, of which two (vaultwarden, zipline) cannot be seeded at all while the catalog rightly closes their sign-up (see R-624). **-- 2026-09-23, the RomM lesson is IN THE HARNESS:** `upgrade-test.py` v2 runs every edge that read back under light load for `--soak` seconds (default 600) and reads the kernel's `oom_kill` counter host-side; a kill or restart turns `proven` into `failed`, a peak over 80 % of the limit adds `memory_tight`. **Red-proof:** romm 5.0.0 → 5.3.0 on the template AS PROMOTED (512M, four workers) — seeded, migrated, read back, and then **OOM-killed at +76 s** under four light callers, verdict `failed` (Docker's own OOMKilled read true here; restarts stayed 0, which is why the walk never saw it). Ten minutes is ample for this failure; demo-hp's first kill came at two hours only because nothing was loading it. Evidence `audits/update-rulings-2026-09-23/harness/`. | **READY — rank P2-MEDIUM; owner: CC (scope already ruled, `09` §3 decision 6)** |
| **R-463** | **[P2-MEDIUM] The day the catalog moves `postgres:16` to `17`, ELEVEN apps are affected and the container image will NOT perform the conversion — and nothing anywhere records that.** MEASURED 2026-09-06: 11 of the 53 templates carry PostgreSQL — **8 on `postgres:16-alpine`**, 1 on `postgres:15-alpine`, plus `postgis/postgis:16-3.5-alpine` and Immich's own `postgres:16-vectorchord…` build. **A grep of the whole register for `pg_upgrade`, "postgres major" or "postgresql major" returns ZERO** (confirmed this session, and confirmed again before filing). **WHY IT IS NOT THE SAME PROBLEM AS R-459, and this is the point of the row: the two engines fail in OPPOSITE directions.** MariaDB starts anyway and skips the conversion quietly, which is why R-459 went unnoticed until a harness looked. **PostgreSQL REFUSES TO START on a datadir from an older major** — the official image performs no `pg_upgrade` and exits with a message naming both versions. So the Postgres case cannot hide; it will present as eight apps down at once, on the sync after the catalog moves. **DELIBERATELY NOT MEASURED HERE, and saying so is the scope discipline:** R-459's task was scoped to MariaDB, and measuring the Postgres analogue is its own piece of work with its own venue. **This row exists so the gap is a record rather than a sentence in an audit nobody greps.** What would settle it: one edge on the existing harness (`postgres:16-alpine` → `17-alpine`) on a scratch host, which would also exercise the `engine_state_after` field's Postgres probe end to end — it is written but has never run against a real Postgres major. Owner: **CC.** `audits/SPIKE-r459-mariadb-upgrade-2026-09-06.md` §7 **-- UPDATE NIGHT 2026-09-21:** **Measured 2026-09-21, both halves.** (a) What a household sees TODAY: the guarded Update of `postgres:16-alpine` to `17-alpine` on docmost ended **`failed` in 5.1 s**, the app stopped and held, **the pin naming 17 while `installed_images` still said 16 and nothing ran**, the data intact, and the restore the hold sentence names back in **29.1 s**. The engine's refusal line had to be REPRODUCED independently because `failAndHold` destroyed it (R-621): *FATAL: database files are incompatible with server / DETAIL: The data directory was initialized by PostgreSQL version 16, which is not compatible with this version 17.11.* The datadir was still `16` afterwards, and the same copy started under 16 holding 48 tables as the positive control. (b) The conversion **COSTED** on a real seeded 49 MB / 48-table datadir: `pg_dumpall` **2.6 s / 132 201 B**, fresh 17 plus replay **6.5 s / 48 tables restored**, the app up on 17 saying *Database connection successful*, **the seeded account read back**, total **155.9 s of which ~9 s is engine work**. `pg_upgrade` was NOT run: it needs both majors' binaries in one image and no such image exists here. Full paragraph: `audits/update-night-2026-09-21/24-Q5-postgres-conversion-costed.md`. **-- RULED 2026-09-23 (`09` §3 decision 16):** PostgreSQL majors are converted BY THE BOX as a guarded-update step — save everything from the old engine, start the new one empty, load it back, check. Each of the eleven apps is proven on the test bench before the catalog may move it; the engine-major gate stays until then. | **READY TO BUILD — owner: CC; `09` §6.4; the gate stays until all eleven are proven** |
| **R-464** | **[P3-LOW] MariaDB's entrypoint prints `MariaDB upgrade not required` on an UNSUPPORTED DOWNGRADE, so that line cannot be used as a soundness signal.** MEASURED 2026-09-06. After converting a datadir to `12.3.3-MariaDB` and then starting **11.6** on it, the entrypoint logs, on every start: **`[Note] [Entrypoint]: MariaDB upgrade not required`**. Asked properly, the same engine answers **`FATAL ERROR: Version mismatch (12.3.3-MariaDB -> 11.6.2-MariaDB): Trying to downgrade from a higher to lower version is not supported!`** **The entrypoint compares the datadir's recorded version against its own and concludes there is nothing to DO. That is true, and it is not a statement that the state is sound.** **THIS IS THIS PROJECT'S MOST-REPEATED CLASS, in a new costume** — the same shape as `CLAUDE.md`'s "presence is not success" and as R-443's HTTP 200 over a crash-looping app: a reassuring sentence that answers a narrower question than the one a reader will take it for. **Why it is worth a row rather than a footnote: the obvious cheap instrument for R-459 is to grep container logs for that exact line**, and such an instrument would report "fine" for an unsupported downgrade. **The correct probe is `mariadb-upgrade --check-if-upgrade-is-needed`**, which is what `upgrade-test.py`'s `engine_state_after` now uses. **Also recorded, because it nearly produced a wrong answer here: run without credentials that command returns `ERROR 1045 … FATAL ERROR: Upgrade failed` with exit 1** — an authentication failure wearing the shape of a verdict. Owner: **CC.** `audits/SPIKE-r459-mariadb-upgrade-2026-09-06.md` §5.4 | **READY — rank P3-LOW; owner: CC** |
| **R-468** | **[P3-LOW] THE GOLDEN WAIVER — goldens on a cadence, not per release (operator ruling 2026-09-13).** 25 goldens in 26 days in August, almost one per release, because `golden_currency_gate.py` trips on every release by design and the only honest ways past it were a bake or a declared `--no-verify` (thirteen by 2026-09-01, R-404/R-417). **The ruling: bake WEEKLY, and always before any drill or fresh install.** Every release still raises the FLOOR, so both demo boxes keep getting each release in ~20 s; only the golden — which protects a fresh install and nothing else — moves to a cadence. **The mechanism (built 2026-09-13):** `documentation/tests/golden-waiver. **⚠ CORRECTED THE SAME DAY (R-472): between bakes the floor does NOT carry a release — the hub holds any floor above the vouched golden (publish-train rule 1), so releases between bakes reach the demo boxes only by hand-deploy.**yml`, four lines (`issued`, `expires`, `reason`, `register_row: R-468`), read by the gate. While valid, a golden BEHIND the record makes the gate print a loud ADVISORY and exit 0; when it expires the gate is red again until someone bakes or renews. **The 14-day cap is enforced by the gate, not the runbook** — a longer, undated, unparseable, reason-less or row-less waiver is INCONCLUSIVE (exit 2), never 0 and never silently ignored. **It never covers a golden that is UNRECORDED (R-385)** — that is not a cadence choice. **A dated waiver cannot be forgotten; it just expires** — the difference from R-242's original rule, which recurred the day after it was written. Tests: `scripts/test_golden_currency_gate.py` cases 5–15 (E/F/G/H, a 15-day, absent, unparseable, bad-row and empty-reason waiver each 2; the R-421 decoy — a file saying only `expires` — 2). **This is a PRE-CUSTOMER arrangement: the first external install retires it** (delete the file in that commit). Cadence written into `RUNBOOK-manual-build.md` §4.2 and the `felhom.eu` end-of-session checklist. **Does NOT touch R-242's open half (nothing gates the VOUCH).** | **WATCHING — rank P3-LOW; owner: CC (renew ≤ 14 days or bake); retire at the first external install** |
@@ -806,6 +806,14 @@ class (an image `VOLUME` at an unmounted path) is still live — `immich-server`
| **R-634** | **[P1-HIGH] An app can be RUNNING, HEALTHY and serving while the controller records it as not deployed — and in that state the household cannot remove it through the product at all.** MEASURED 2026-09-22 on guest 9202, controller v0.261.0, on **two independent apps in one night**: `outline` and `sparkyfitness`. **The controller's own words, in order.** `outline` deploy accepted 11:54:47; **`11:55:51 StopStack outline: current state=deploying deployed=true containers=0`** — a stop while the stack is still deploying; `11:56:09 SaveAppConfig: saving /opt/docker/stacks/outline — 5 env vars, **0 encrypted**, 3 sensitive fields` (the two saves before it both read `3 encrypted`); then **`11:57:16` and `12:02:16 Health probe outline: API GET :3000/_health -> 200`** — the app is up and answering its own health endpoint; and **`12:02:19 StopStack outline: current state=running deployed=false containers=3`**. Three containers, state `running`, health 200, **`deployed=false`**. The remove then answers **`RemoveStack outline: state=not_deployed, deployed=false, orphaned=false, deploying=false` -> `[ERROR] Remove failed for outline: stack "outline" is not deployed`** for BOTH the remove-with-data and the remove-keeping-data call, while `ScanStacks` goes on finding the stack every ten seconds. `app.yaml` survives with `desired_state: stopped` and an **empty `installed_images`**. `sparkyfitness` produced the identical shape 13 minutes earlier. **WHAT THIS COSTS A HOUSEHOLD:** an app that works is invisible to the product as an installation — no badge, no update, no backup selection, and **no way to delete it**; the only exit is a shell. It is the mirror of R-633 (there the record is gone and the container remains; here the container is fine and the record is gone) and it is the **worse** of the two, because the app is serving customer traffic the whole time. **THE MECHANISM IS NOT DIAGNOSED, and this row says so rather than guessing.** What was tried: the controller's full container log for both apps (the sequence above), `app.yaml` on disk, `docker ps -a`, and `GET /api/stacks/<n>`. What was NOT done: reading `runComposeDeploy`'s pin-write path — this was an unattended run and the brief forbade product code. **The one discriminator worth running first:** both apps were walked while two other walks ran concurrently, and `POST /api/backup/run` is box-wide, so a backup or restore for a NEIGHBOURING app was in flight. A serial re-walk is queued tonight; if it reproduces alone, concurrency is not the cause. **A THIRD THING THE SAME LOG SHOWS, recorded here because it is one line away:** at `11:55:57` the health probe dialled **`http://outline-postgres:3000/_health`** — during startup, with the exactly-named container not yet running, `findProbeContainer`'s PREFIX fallback latched onto the POSTGRES sidecar and probed port 3000 on it. Transient, and it resolved once `outline` came up, but it is the same function R-630 is about. Evidence: `audits/the-28-2026-09-22/apps/half-state-outline-sparkyfitness.txt`. **THE SERIAL RE-WALK WAS RUN THE SAME NIGHT AND IT SPLITS THIS ROW IN TWO — recorded here rather than left as the first reading.** Walked again one at a time, with no other walk running: **`outline` deployed normally and removed clean**, and **`crafty-controller` deployed normally, updated `4.10.7 → 4.11.0` to `done`, restored and removed clean.** So for those two the half-state did NOT reproduce alone, and concurrency — a box-wide `POST /api/backup/run` or a restore in flight for a NEIGHBOURING app — is implicated rather than the deploy path itself. **`sparkyfitness` reproduced EXACTLY, alone, in 534 s**: deploy accepted, never reached `deployed` with a pin, `app.yaml` left with `desired_state: stopped` and an empty `installed_images`, and **both remove calls refused with `stack "sparkyfitness" is not deployed`.** **So the row stands, at one reproducible app instead of three, and the honest split is:** (a) `sparkyfitness` has a deploy that does not finish and leaves a record the product cannot clear — reproducible, P1; (b) under concurrent work the same unremovable half-state can be reached by apps that are otherwise fine, which is the more alarming half because those apps were **running, healthy and serving** while recorded as not deployed. **Neither half is diagnosed** — `runComposeDeploy`'s pin write was not read, because the brief forbade product code. **THE BOUNDED HALF IS FIXED in v0.262.0; the MECHANISM IS STILL NOT DIAGNOSED, and this row stays open for it.** `RemoveStack` refused on `!stack.Deployed` — a FLAG — while the machine plainly had containers, a compose file and an `app.yaml`. It now asks whether anything EXISTS (`halfStateEvidence`): containers, a compose file, or an `app.yaml` are each enough, and it removes what exists with every other fence unchanged (R-442 stays fail-closed). **The household must always be able to remove what the box shows them** — that iLine truncated
| **R-635** | **[P1-HIGH] `romm 5.3.0` does not fit the memory the template gives it, and the guarded Update called that a success — the app has been OOM-crash-looping on demo-hp for six hours at ~500% CPU.** FOUND 2026-09-22 17:37 **because the operator heard the fans**, which is the only reason it was found at all. `romm` was promoted `5.0.0 -> 5.3.0` on the live catalog that morning (`audits/PROBE-FIX-2026-09-22.md`) after the edge was PROVEN on scratch guest 9202, and the guarded Update was then pressed on demo-hp guest 9201 at **09:08:22Z**, reaching **`done` in 74.8 s** with the app `running`. **It ran clean for two hours.** The first worker kill is at **11:09:20Z**; by 15:38Z there had been **4,530** of them — `Worker (pid:…) was sent SIGKILL! Perhaps out of memory?` — with `docker inspect` reading **`OOMKilled: true`**, the container pinned at **457 MiB of its 512 MiB limit**, and `docker stats` showing **499.51% CPU**. The host's load average sat at **5.2 while otherwise idle**. `rq_cron` is killed and restarted every few seconds in a permanent storm. **THREE THINGS THIS ESTABLISHES, and the third is the one that changes how promotions are judged.** (1) The template's own comment says *`RAM: ~300MB (mem_limit: 1024M total — romm 512M + mariadb 384M + redis 128M)`*; **5.3.0 needs more than 512M and the template was not re-sized when the version moved.** A version move is not only an `image:` line. (2) **The update reported `done` and the app reads `running`**, because nginx answers `GET /` with 200 while the gunicorn workers behind it are being killed — a THIRD variant of the R-618/R-630 theme: the probe is right, the port is right, and the answer is still a false green. (3) **THE ALARM DID FIRE, AND MY FIRST WRITE-UP OF THIS ROW SAID IT DID NOT — CORRECTED 2026-09-22 BY THE OPERATOR, WHO PRODUCED THE MAILS.** The controller HAS an OOM detector (`main.go:821`, `[deadapp] romm: container romm was OOM-killed`, evaluated every 30 s), it emits **`app_oom`** (`notifier.go:726`), the hub allow-lists it (`dispatcher.go:636`) and dispatched it to the OPERATOR channel as **SENT** — `admin@felhom.eu` received **`[Felhom] ⚠️ demo-hp: app_oom`** at **11:09 CEST** and again at **17:48 CEST**, each carrying the app name, the Hungarian sentence and the dashboard link. The CUSTOMER channel is **SKIPPED**, which is correct. **I asserted an absence without looking at the instrument** — the hub's own Events and Notifications tabs show all of it — and I did it by reasoning from a memory note (`lxc-docker-oom-signals-unreliable`, R-528) instead of reading the hub. **That is R-628's shape exactly, four days on, and from the same hand.** **WHAT IS ACTUALLY WRONG, and it is narrower and real:** `notifier.go:715-726` keys the alarm on `container|startedAt` in an `oomSeen` map and emits **once per container lifetime**. So **six hours of continuous thrashing — 4,530 worker kills — produced exactly ONE mail**, severity `warning`, never escalating, while the app went on reading `running`. **A six-hour storm is indistinguishable from a single transient kill.** The hub's App Telemetry panel did carry the magnitude (RomM: **5,023 errors, 632 warnings**, peak 855 MB) but nothing turns that into a second, louder signal. So the fix worth having is not a detector — there is one — but an ESCALATION: a warning that repeats for hours should stop looking like a warning that happened once. **AND THE LESSON FOR R-462's METHOD, which is the real cost:** every `proven` verdict in the update night and in the twenty-eight measures the app for the **minutes of the walk**, not for a day of running. `romm` passed its walk, was seeded, read back and restored — and broke two hours later. **`proven` currently means "the update applied and the data survived", NOT "the new version runs".** Needs: decide between rolling the catalog back to 5.0.0 and raising romm's `mem_limit` (measured, not guessed); and a soak longer than a walk before any future promotion. **FIXED AND MEASURED 2026-09-22, in two steps, and the FIRST step was still a guess.** *Step 1 (operator's choice):* the limit was raised 512M → **768M** (`app-catalog-felhom.eu@886956d`). It slowed the kills from ~12/min to ~7/min and **stopped nothing** — 37 SIGKILLs in five minutes, `OOMKilled` still true, and the cgroup's own `memory.peak` read **exactly 768 MiB**: it hit the new ceiling and died there. *Step 2, from a MEASUREMENT instead:* the per-process RSS inside the container reads **~216 MiB per warm uvicorn worker**, so the image's default of four workers plus the master needs **~882 MiB** before nginx and the job runner — more than any sensible limit for this box. **The lever was in the image all along:** `/init:143` runs `--workers "${WEB_SERVER_CONCURRENCY:-4}"`. **Four workers is a SERVER default on an appliance serving one household.** Setting **`WEB_SERVER_CONCURRENCY=2`** (`app-catalog-felhom.eu@f4eb94f`, limit left at 768M) and applying itLine truncated
| **R-636** | **[P2-MEDIUM] An app that has been out of memory for six hours sends the same single warning an app that hiccuped once sends.** FOUND 2026-09-22 when the operator produced the alarm mails I had wrongly written off as absent (R-635). **The detector is correct and works.** `main.go:821` re-checks every 30 s and logs `[deadapp] romm: container romm was OOM-killed`; `notifier.go:726` emits **`app_oom`** at severity `warning`; the hub allow-lists it (`dispatcher.go:636`) and delivered it to the OPERATOR channel — two mails, 11:09 and 17:48 CEST, each naming the app and linking the dashboard. CUSTOMER is SKIPPED, correctly. **The defect is the SHAPE of the signal, not its absence.** `notifier.go:715-724` keys on `container|startedAt` in an `oomSeen` map and returns early on a repeat, so the alarm fires **once per container lifetime**. romm's 09:08 container produced **4,530 worker kills over six hours and exactly one mail**; the 15:47 container produced one more. Severity never escalates, the app goes on reading `running`, and `08` §4 rightly does not put it in `IsDownState`. **So a six-hour storm that pins five cores is indistinguishable, in the operator's inbox, from one transient kill at 3 a.m.** — and the operator, who had two correct mails, still found the fault by hearing the fans. **The once-per-lifetime rule is right in itself** (it is what stops a crash loop from mailing 4,530 times, which is R-629's lesson) — what is missing is the second, louder signal when the same key keeps re-firing. **The magnitude IS already collected:** the hub's App Telemetry panel read **RomM: 5,023 errors, 632 warnings, peak 855 MB against a 1280M limit** while the same panel showed every other app at 0. Nothing turns that into an event. **Candidate shapes, none chosen here:** escalate `app_oom` to `error` when the same `container|startedAt` key re-fires past a threshold; or a periodic digest for a key still firing after N minutes; or let App Telemetry raise its own event when an app's error count crosses a bound. All are hub/controller code. Evidence: the operator's screenshots of the Events, Notifications and App Telemetry tabs plus the two mails; `felhom-controller/controller/internal/notify/notifier.go:715-726`. | **OPEN — P2; owner: CC; product code, so not fixed unattended** |
| **R-637** | **[P2-MEDIUM] BUILD THE UNDO (`09` §3 decision 15) — the eight things the 2026-09-23 spike says the product must add before a failed update can put itself back.** SPIKED BY HAND 2026-09-23 on 9202, three real migrating edges each made to fail a deliberately wrong probe: docmost (PostgreSQL) and romm (MariaDB) — whose OLD versions REFUSE the migrated data (*corrupted migrations …*, *Can't locate revision …*) — and vikunja (SQLite in a volume), whose old version starts. **The undo worked on all three**: data written before AND after the backup read back through each app's own front door, ≈16 s / ≈38 s / ≈1 s after the failing health wait. **The list:** (1) keep the pre-update copies — compose, applied, pin AND the old `.felhom.yml` — until the undo is over (`failAndHold` deletes the first three, the fourth was never kept, R-639); (2) the undo inside `failAndHold`: pin back → DB up alone → validated load → start → health with the OLD probe → `undone`, else HOLD; (3) empty-then-load atomically, because the loader the product has cannot undo a migration (R-638); (4) check the copy's completion marker first (R-640); (5) a volume copy at safety-dump time for apps with no database server (R-641); (6) files: nothing measured touched them — a step that does says so through decision 13's *files may change* mark; (7) the undo's success is the probe, never the Start's return (R-642); (8) remember the failed step so the caller never re-presses it. `09` §6.1a, §6.4 part 1 (4 evenings). Evidence: `audits/update-rulings-2026-09-23/README.md`. | **READY — owner: CC; `09` §6.4 part 1, returns to the operator for go/no-go** |
| **R-638** | **[P2-MEDIUM] The product's database loader cannot replay a copy over a NEWER schema: on PostgreSQL it FAILS, on MariaDB it leaves the newer version's tables behind.** MEASURED 2026-09-23 on 9202. `ImportDump` (`appbackup/dbdump.go:719`, `psql -v ON_ERROR_STOP=1 --single-transaction`) replays a `pg_dump --clean --if-exists` file over the live database. After docmost 0.95.0 → 0.96.0 migrated, the replay of the pre-update copy was refused in 0.40 s, rc 3: *cannot drop constraint workspaces_pkey on table public.workspaces because other objects depend on it / DETAIL: constraint oauth_clients_workspace_id_fkey …* — the new version created six tables whose foreign keys point at old ones, and `--clean` only drops what the dump knows. Database unchanged (the transaction rolled back). On MariaDB (`mariadb-dump`, `FOREIGN_KEY_CHECKS=0`) the same replay after romm 5.0.0 → 5.3.0 returned rc 0 in 1.25 s and left **12 base tables** of the new version behind; RomM 5.0.0 happened to ignore them. **What worked:** `DROP SCHEMA public CASCADE; CREATE SCHEMA public;` + the dump in ONE transaction — rc 0 in 1.38 s, every table, index and extension back. **Why this is a row of its own and not only part of R-637:** the SAME loader backs shipped paths — `rollbackSafetyDump` (off-site restore's undo) and the dump replay of the restores — so **any restore of a copy taken BEFORE an update that migrated, replayed over the migrated database, may fail the same way. NOT MEASURED:** whether the unit restore the hold sentence names does this (it also carries the data VOLUME tar, which may make the replay moot). That is the measurement owed, on 9202, before anyone relies on it. Evidence: `audits/update-rulings-2026-09-23/README.md` Part 1, `docmost-45`, `romm-44`. | **OPEN — P2; owner: CC; measure the named restore after a real schema migration first** |
| **R-639** | **[P3-LOW] After a failed update is held, the previous definition survives only in the recovery unit — `failAndHold` deletes the journal's copies, and the old `.felhom.yml` was never kept.** READ FROM SOURCE and CONFIRMED 2026-09-23: `failAndHold` (`stacks/update.go:766`) ends with `clearJournal` + `removePreUpdateCopies`; in all three spike cases `ls <stack>/*pre-update*` counted 0 after the hold. The old definition was taken from the unit's `compose/` directory each time — present only because a backup preceded the update. The old `.felhom.yml` matters as much: its probe is the one the OLD version answers, and the catalog's copy flows to the app regardless (§5.4). Needed by R-637. Evidence: `audits/update-rulings-2026-09-23/README.md`, `docmost-41`, `romm-40`. | **OPEN — P3; owner: CC; lands with R-637** |
| **R-640** | **[P2-MEDIUM] A TRUNCATED PostgreSQL copy loads with exit 0 into an EMPTY database — and `ValidateDump` would accept it.** MEASURED 2026-09-23 on 9202 in a scratch database beside docmost's: the first half of the 135 816-byte undo copy, loaded with `psql -v ON_ERROR_STOP=1 --single-transaction`, returned **rc 0** and left **42 tables and 48 migration rows — and 0 users, 0 spaces, 0 constraints, 0 indexes**; psql treats end-of-file inside a `COPY` as end of data and commits. The whole copy ends with `-- PostgreSQL database dump complete` (plus a `\unrestrict` line); the truncated one does not. `ValidateDump` (`appbackup/dbdump.go:415`, read from source) checks the header and a `CREATE TABLE` only. **An undo or a restore fed a truncated copy would report success, start the app on an empty database, and pass a health check.** MariaDB's truncated load failed (rc 1) but NON-atomically — half the tables already replaced; its copy carries `-- Dump completed`. Fix: check the engine's completion marker before any load. Evidence: `audits/update-rulings-2026-09-23/README.md`, `docmost-46`, `docmost-47`, `romm-44`. | **OPEN — P2; owner: CC; lands with R-637, also guards the restore paths** |
| **R-641** | **[P2-MEDIUM] An app with no database server has NO last-second copy — the update's safety dump is a no-op — so if its old version refuses the migrated data, everything written since the last backup is lost by any route back.** MEASURED 2026-09-23 on 9202 with vikunja (SQLite in a volume): the controller logged *update safety dump for vikunja: the app has no database — nothing to copy (no-op)*. Vikunja 2.3.0 happens to start on 2.6.0's migrated SQLite, so nothing was lost; the fallback was measured anyway — the tier unit's volume tar put back in 0.86 s brought back the data written before the backup and **NOT the project and attachment written after it**. The fix is a tar of the data volumes at safety-dump time (2.2 MB here). Needed by R-637. Evidence: `audits/update-rulings-2026-09-23/README.md`, `vikunja-40`..`-44`. | **OPEN — P2; owner: CC; lands with R-637** |
| **R-642** | **[P3-LOW] `POST /api/stacks/{name}/start` answers 200 *Stack … start completed* while the app is crash-looping.** MEASURED 2026-09-23 on 9202 twice: docmost 0.95.0 and romm 5.0.0 started on data their newer versions had migrated — both refused and restarted in a loop (`Restarting (1)`, front door 404) behind a 200. The same false-green class as R-443 (closed for the Update) and R-635, on the Start. It matters now because an undo (R-637) must never read the start's return as success. Evidence: `audits/update-rulings-2026-09-23/README.md`, `docmost-43`, `romm-43`. | **OPEN — P3; owner: CC** |
| **R-643** | **[P2-MEDIUM] The ruled chain leaves the automatic update leg AT MOST 15 MINUTES a night.** FOUND 2026-09-23 while writing the build plan for `09` §3 decision 11 (*updates after the off-site copy, before the full-system backup*). The off-site leg starts at W+105m (`cmd/controller/main.go:961`) and the full-system backup's gate opens at W+2h (`quiesce/quiesce.go:656`, span to W+6h); the legs are clock-scheduled, not chained. One step takes ~1 min when it works and ~2–6 min when it fails and is undone. Options and the recommendation (the full-system backup waits for the leg inside its own window; the leg stops starting steps at W+5h) are in `09` §6.4. | **WAITING-ON-OPERATOR — `09` §6.4's one open point; owner: CC once answered** |
| **R-644** | **[P3-LOW] `gokapi` on scratch guest 9202 is crash-looping — 329 restarts by 2026-09-23 07:51 UTC, *password does not appear to be a SHA-1 hash* — and the controller still lists it deployed.** OBSERVED at the start of the 2026-09-23 session, not caused by it. The twenty-eight walk's teardown (2026-09-22) removed a `gokapi` container left by R-633 by name; a `gokapi` is running again, recorded `deployed: true`. Not investigated (scope). Likely the R-633/R-634 shape — a restore-then-remove race leaving a record — and a scratch-box fact, not a customer one; filed so the next drill does not read it as its own. | **OPEN — P3; owner: CC; investigate before the next drill on 9202** |
<!-- DUE-CHECKS-BEGIN — machine-readable. Parsed by scripts/due_checks_gate.py.
One row per dated check. The R-number must have a row above. Dates are UTC.
Clearing a row means the check was DONE and its result recorded in that R-row —