docs(report): lifecycle implementation, the v0.158.1 defect, and the R-55 live leg

This commit is contained in:
2026-07-21 16:36:14 +02:00
parent 0fbd272ad2
commit 4aa2ce4b61
+108 -95
View File
@@ -1,122 +1,135 @@
# REPORT — R-55 boot-gate stop-honesty (TASK-E Part 2), 2026-07-21
# REPORT — app lifecycle states + R-55 live leg (TASK-F Parts 1 & 3), 2026-07-21
**Versions: v0.157.0 (R-55) + v0.157.1 (.gitignore anchor). Baseline `83f20c8` (v0.156.0), clean tree.
v0.157.0 is BUILT, PUSHED and DEPLOYED to guest 9201.**
**Versions: v0.158.0 (lifecycle) + v0.158.1 (a defect I shipped in 0.158.0 and caught live — read §4).
Baseline `a96226a` (v0.157.0), clean tree. Deployed to guest 9201 and live-verified.**
## Scope correction — this part was specified against the wrong repo
## 1. What shipped — app lifecycle
TASK-E placed R-55 in `felhom-agent` (Parts 12 → felhom-agent) and separately forbade controller
changes ("felhom-controller untouched", "no controller changes at all"), with a supervised-lite leg
described as an *agent binary* deploy to felhom-pve.
`.felhom.yml` gains an optional top-level `lifecycle:`. Absent/empty ≡ `available`, so all existing
templates are untouched.
**R-55 is not in the agent.** It is `felhom-controller/controller/internal/web/intermediary.go`
the in-guest boot bind gate. The agent's only related code is host-side `ReassertGuestBinds`, which
re-adds `pct set -mpN` binds and never touches apps. The GO ruling and the scope fence therefore
contradicted each other; the fence was evidently written believing the gate lived in the agent.
| state | offered for new installs | shown to someone already running it |
|---|---|---|
| `available` | yes | nothing |
| `hidden` | **no** | nothing — "we stopped offering this" is not their problem |
| `abandoned` | **no** | „Nem karbantartott" badge + a permanent notice on the app page |
**Raised as a STOP rather than resolved unilaterally**, because it changes which repo ships and which
artifact deploys. **Operator ruling 2026-07-21: implement in the controller AND deploy to 9201.** Done.
**A deployed instance keeps full function in every state.** That is the whole design: the earlier
`retired/` directory move un-offered the app but would ALSO have made the orphan detector see the
template as gone for anyone running it — flagging a working install `Elavult` and offering a Törlés
button. Withdrawing an app must never take a working app away from a customer.
## The defect
Three design points worth carrying past this feature:
`shouldRecreateOnBoot` keyed on `Deployed && HDD_PATH && drive-present`. `Deployed` is a
deploy-lifecycle flag that stays true across a Stop, so the gate could not tell "the guest went down
under this app" from "the customer switched this off" — and resurrected both. Proven live by STOP-1's
R-52 leg, which was designed to prove the opposite: immich, stopped from the UI seconds earlier, came
back running after `pct reboot 9201`.
- **The deploy gate is server-side and fail-CLOSED**, before any mutation, with the ruled message
„Ez az alkalmazás jelenleg nem telepíthető." Hiding a button is not a gate — a stale link, a
bookmarked deploy form and a direct POST must all be refused. A second check in
`stacks.DeployStack` covers callers that do not route through the API.
- **Unknown values fail OPEN** (→ `available` + one WARN) — deliberately the opposite posture. A
typo, or a state from a catalog newer than this controller, must never silently pull a working app
out of every customer's list. Both paths read the same `EffectiveLifecycle`, so they cannot disagree.
- **Lifecycle never reaches orphan detection.** `getCatalogTemplateSlugs` keys on directory + compose
presence only. Red-proofed.
The operator's rationale for ruling this GO is worth keeping attached to the fix: an unconditional
recreate can **drown the node** when apps were deliberately stopped to free resources for others. This
is not only about honesty; it is about not fighting the customer's own capacity management.
Badge plumbing is generic (`MetaBadge` + the `meta_badge` partial + a `lifecycleBadge` funcmap
entry) so R-56's difficulty labels need only a sibling funcmap function — no new markup or CSS.
## Spike-lite: a truthful source already existed — nothing was invented
## 2. Live validation — endpoint-level, on 9201
The task required stopping rather than inventing state if no honest running-at-shutdown signal existed.
It does exist, and R-52 already trusts it:
- `AppConfig` carries **no** run-state. `Deployed` is lifecycle, not intent.
- `Stack.Containers` comes from `docker ps -a` (so `Exited` containers **count**), and docker's own
on-disk container records survive a reboot.
- A UI Stop is `compose down`, which **removes** the containers. Nothing else leaves a deployed app at
zero containers.
That is exactly `bootrecon.isBootOrphan`'s `existing-Exited vs absent` distinction. The gate now
additionally requires `len(Stack.Containers) > 0`.
**What deliberately did NOT change: container STATE is still not a filter.** That half of the original
design is load-bearing — a `State != stopped` filter misses an app docker has not auto-restarted yet,
or one stuck `Exited` on a create-time bind failure with `RestartCount=0`, which is the bug the boot-id
path exists to fix. `state` and `hasContainers` are different questions ("is it up right now" vs "does
docker still have records of it") and only the second survives a reboot as a statement of intent.
`TestShouldRecreateOnBoot` now pins both axes at once; they pull in opposite directions, which is the
entire difficulty of this gate.
**Ordering trap, found and handled:** the evidence is sampled into the `bootStack` snapshot **before**
any recreate runs, because `recreate` itself calls `StopStack` (`compose down`) and destroys it.
**No regression to the drive-absent gate:** apps it stopped are also at zero containers, so this path
now skips them — correctly. They are recorded in `StoragePath.StoppedStacks` and restarted by
`ReconcileDriveGates`' `Return` branch on the same `driveGateLoop` tick. Checked, not assumed.
**Honoured Stops are observable.** `leftStopped` is counted and logged separately from `skipped`, at
INFO. Conflating them would fire a WARN about a missing drive bind for an app behaving exactly as
asked — and a silent correct path is how an inert seam hides.
**No new privileged verb**, as required: `StopStack`/`StartStack` are plain `docker compose` execs in
the controller's own container, and the boot path's only agent calls are read-only (`agent.Disks`).
The agent sudoers/wrapper set is untouched — nothing in it relates to app start/stop.
### Red-proof — RUN
Dropped `hasContainers` from the predicate:
The catalog change arrived through the **normal git-sync**, not a hand-copy:
```
--- FAIL: TestShouldRecreateOnBoot
R-55 customer-stopped drive app (zero containers): shouldRecreateOnBoot = true, want false
R-55 gate-stopped app (zero containers; Return branch owns it): shouldRecreateOnBoot = true, want false
--- FAIL: TestRecreateDriveBackedApps_HonoursCustomerStop
recreated=[romm immich], want exactly [romm] — a customer-stopped app must never be restarted
$ grep -m1 '^lifecycle:' /opt/docker/stacks/plant-it/.felhom.yml
lifecycle: abandoned
```
`recreated=[romm immich]` is the live defect by name. Fix restored → green; `git diff` clean.
| check | result |
|---|---|
| plant-it on the Alkalmazások page | **0 occurrences** (control app `bookstack`: 10) |
| direct `POST /api/stacks/plant-it/deploy` | **HTTP 409** `{"ok":false,"error":"Ez az alkalmazás jelenleg nem telepíthető."}` |
| `/apps/plant-it` | 200, carries the notice, **no Telepítés button** |
| `/apps/bookstack` (control) | 200, no badge, unaffected |
## Gates and deploy verification
Method: the exact endpoints the UI calls, authenticated as the customer (no browser on DooPlex).
`go build ./... && go vet ./... && go test ./...`**green**, whole module.
## 3. Red-proofs — all five run
| # | pre-fix shape restored | observed failure |
|---|---|---|
| 1 | remove the API deploy gate | wiring test: *"deployStack never calls Meta.CanInstall() — the lifecycle gate is INERT"* |
| 2 | drop `st.Deployed \|\|` from the catalog filter | `"plant-it-run" must be listed`, `"oldapp-run" must be listed` — a customer's running app vanishes from their own page |
| 3 | remove the badge line from stacks.html | *"a deployed abandoned app must carry the Nem karbantartott badge"* |
| 4 | make orphan detection lifecycle-aware | `catalog set = map[bookstack:true]` — both withdrawn apps read as orphans |
| 5 | restore the pointer receivers (§4) | `can't evaluate field IsAbandoned in type interface {}` on every lifecycle value |
The wiring test walks the **AST**, not `strings.Contains`, because a commented-out call still
contains the string; it also asserts the gate precedes `DeployStack`, since a gate that fires after
the mutation is not fail-closed.
## 4. A defect I shipped in v0.158.0 and caught within the hour
**`/apps/<slug>` returned HTTP 500 for EVERY app**, not only withdrawn ones.
`EffectiveLifecycle` / `CanInstall` / `IsAbandoned` were declared with **pointer receivers**.
`appDetailHandler` passes `data["Meta"] = found.Meta` — a `stacks.Metadata` VALUE inside a
`map[string]interface{}` — and html/template cannot call a pointer-receiver method on a
non-addressable value:
```
gitea.dooplex.hu/admin/felhom-controller:0.157.0 | Up (healthy)
executing "app_info" at <.Meta.IsAbandoned>: can't evaluate field IsAbandoned in type interface {}
```
**Not an inert seam** — the new code is verified present in the *running* binary, not just the image:
`grep -a -c "stopped them on purpose" /usr/local/bin/felhom-controller``1` inside the live
container. This check exists because three inert-seam defects shipped fully-green in three days.
Fixed in v0.158.1 with value receivers and the reason recorded at the declaration so it is not
"tidied" back.
### The live reboot leg is NOT done, and that is a deliberate stop
**Why every gate missed it, which is the transferable part.** It compiles. `go vet` is silent. All
of v0.158.0's tests passed — because none of them rendered `app_info`. The catalog-page tests
exercised the *funcmap* route (`lifecycleBadge .Meta`), which takes a value and works with either
receiver. **A template method call is only ever checked when that template actually runs**, so a
render test is not optional coverage for template-facing methods — it is the only coverage.
Proving R-55 end to end needs a drive-backed app stopped from the UI followed by `pct reboot 9201`.
9201 has two drive-backed apps (calibre-web, immich) on `hdd_1`, so the fixture is available — but the
demo box is **at a remote site until ~2026-08-02**, and rebooting it unprompted is not a call CC
should make (the 2026-07-20 dhclient incident cost 1h15m from a smaller action). **One operator-present
STOP away.** Until then R-55 is *shipped and wired*, not *proven live*; the capability-map row says so
in those words.
`TestAppInfoRendersForEveryLifecycle` now renders the real template through the production tree with
the handler's exact data shape — `"Meta"` as a value in a `map[string]interface{}`, deliberately not
a pointer, because a pointer would hide the bug the test exists to catch.
## v0.157.1 — `.gitignore` anchor (TASK-E Part 5)
I found this while live-validating, not from the suite. That is the honest sequence.
`controller/.gitignore` carried a bare `controller`, which git matches against **directories** too, so
it also matched `cmd/controller/`. Two opposite failure modes, both manufacturing inert seams:
ripgrep silently skipped `cmd/controller/main.go` (a search for a setter's caller returns nothing and
reads as "unused"), and new files there needed `git add -f` or were never committed.
## 5. Part 3 — R-55 live leg, PROVEN
Anchored to `/controller` + `/controller.exe`. Verified **both** directions: `git check-ignore` no
longer matches `cmd/controller/main.go` and `rg` now finds it; a built binary at the module root is
still ignored. `git ls-files` confirms nothing had been hiding untracked behind the old rule.
Operator-present, 2026-07-21. Fixture on 9201 (guest clock = UTC, host = CEST):
## Observations
| time (UTC) | event |
|---|---|
| 14:28:52 | **immich STOPPED through the real UI endpoint** `POST /api/stacks/immich/stop``compose down` |
| — | pre-state verified: immich **0 containers**, calibre-web 1 running (drive-backed), bookstack 2 running (not drive-backed) |
| 14:29:14 | `pct reboot 9201` |
| 14:29:28 | `[gate] boot 1784525102-11906045: waiting (≤2m0s) for live drive bind(s) [/mnt/felhom-drives/hdd_1]` |
| 14:29:28 | `[gate] … live bind confirmed — recreating drive-backed app calibre-web (state=stopped) onto /mnt/felhom-drives/hdd_1` |
| 14:29:29 | `[gate] … 1 drive-backed app(s) left stopped — zero containers means the customer stopped them on purpose` |
| 14:29:32 | `[bootrecon] Boot reconciliation: no boot-orphaned apps (nothing to start)` |
- `TestShouldRecreateOnBoot`'s old comment argued explicitly that a state filter would be a
regression — correct, and still true. The comment is now expanded to say why `hasContainers` is a
*different* axis, so the next reader does not "fix" it back.
- The gate calling `StopStack` before `StartStack` on every recreate is what erases the container
evidence, and is also why R-52's boot-orphan sweep only ever sees non-drive-backed apps. R-52's
practical scope on a real box remains narrower than its brief assumed; unchanged by this fix.
End state: **immich 0 containers / stopped**, calibre-web 1 running, bookstack 2 running. **Zero
alerts** (the only `app_start_failed|deadapp` grep hit was the scheduler registering its job).
Reboot → steady state in **~15 seconds**.
**immich is absent from the recreate list.** Hours earlier, the identical fixture brought it back
RUNNING — that is the defect R-55 fixed, and this is the same experiment with the opposite result.
The `left stopped` INFO line fired in production for the first time, so the honoured path is
observable rather than silent.
## 6. Gates
`go build ./... && go vet ./... && go test ./...`**green** (both versions). All five template
gates pass (`template_id`, `emoji`, `mojibake`, `app_row_dedup`, `native_confirm`).
## 7. Observations
1. **I pushed a commit that does not build.** A `git stash`/`stash pop` used to compare
`reuse_refs_check.py` output silently discarded the staged index, so the first lifecycle commit
landed only the new test files. Fixed by a follow-up commit rather than a force-push (rewriting
pushed history is worse than one bad commit in the log). **Lesson: never `git stash` mid-staging.**
2. `REUSE.md` cross-repo citation drift is pre-existing and improved: 8 broken paths → **5**. I fixed
the three that were plain prefix errors; the rest cite abbreviated paths the checker resolves
against the repo root.
3. The `hidden` state currently has no user in the catalog. It is implemented and tested, but only
`abandoned` is exercised end to end.