felhom-controller/TEST-REPORT.md

TEST-REPORT — Test campaign #2 (broader apps, edge cases, break-and-recover) — N100 / guest 9201

Run start: 2026-06-22 (CC, unattended). Operator: Viktor (away). Scope: push past campaign #1 — more apps, edge cases, chaos break-and-recover. Storage lifecycle now unblocked (ctrl v0.74.0 fixed the agent leak). Only code changes: trivial catalog HC/config-seed fixes. Legend: PASS / FAIL / SKIPPED(reason). Evidence raw (docker ps / ss / curl / screenshot). (Campaign #1 report preserved in git history + felhom.eu/documentation/tests/unattended-test-campaign-2026-06-22-*.md.)

---

Phase 0 — Baseline + safety net + leak-fix gate — PASS (gate OPEN)

• Baseline: ctrl v0.74.0 healthy, agent v0.39.0; pct list = 9201 running (+ 9001/9999 stopped, pre-existing); 25 containers, 0 unhealthy; rootfs / 4%, /var/lib/docker 8%, drives felhom-flash 1% / felhom-usb 1%.
• Master safety net + DR source: PBS backup felhom-pbs:backup/ct/9201/2026-06-22T16:13:44Z, success=true, crash-consistent; PBS verify shows the snapshot present.
• Leak-fix gate (the #1 unblock): /api/disks lists felhom-flash/felhom-usb/local; /api/host-metrics returns cpu_percent; controller→agent idle sockets to :8443 flat at 2 before and after a 60-call burst. Agent path fully working. Gate to chaos/destructive phases: OPEN.

Phase 1 — Broader app sweep + edge cases — PASS (with noted non-triggerables)

Check	Result	Evidence
Deploy navidrome (HDD-bound, Go) via GUI	PASS	GUI deploy → 3-step panel → running/healthy in 8s; :4533 → HTTP 302 (serving); HDD binding …/felhom-flash/userdata/media/music → /music
Pattern coverage	PASS (cited)	JVM=komga (healthy), Python+DB=immich/paperless (healthy), Go=navidrome/gitea, HDD-bound=navidrome/romm/immich, optional-config=romm
Edge: rapid deploy→stop→redeploy race	PASS	stop → immediate update → navidrome Up (healthy), app.yaml deployed:true, API state=running deployed=true — no stale button / no Deployed-flag race
Edge: slow-DB-migration start_period	PASS (cited)	mealie deployed healthy (start_period 60s covers migration — no premature unhealthy flap; lesson #7)
Edge: missing required config → graceful signal	PASS (cited #1)	campaign-1 glance: crash-loop with a clear logged error (open /app/config/glance.yml: no such file), controller marks it — not a silent hang
Edge: mem-gate hard-block	PARTIAL	the deploy-page memory projection renders correctly (e.g. navidrome 4065→+50 MB / 11904); a RAM hard-block isn't naturally triggerable (no single catalog app's request > ~7.6 GB free without mass-filling RAM). Disk-headroom hard-refuse (GetDockerVolumeHeadroom) is tested in Phase 3/C4.
Edge: Host()/port overlap	N/A (noted)	each catalog app has a unique subdomain by construction; an overlap isn't naturally producible without editing a template — not triggered

navidrome left deployed (used as the drive-backed app for Phase-3 C6); removed at cleanup.

Phase 2 — Storage lifecycle (the campaign-#1 blocker, NOW UNBLOCKED) — PASS (destructive-format skipped, documented)

Agent path works (v0.74.0), so these ran through the real controller→agent pipeline.

Check	Result	Evidence
Scan / observe / health / FS / role / durable-id	PASS	/api/disks: 5 targets — felhom-flash (user-data, uuid:81a26531), felhom-usb (user-data, uuid:da9e7089), local/local-lvm (system), felhom-pbs (backup) with capacities + data_bearing
Label edit → revert (live)	PASS	felhom-usb label → "TEST-LABEL-c2" (log "Storage label updated") → reverted to exact UTF-8 original "Tárhely (felhom-usb)" (byte-verified). [my shell mangled the accent on first revert; corrected via %C3%A1]
Disconnect → reconnect (reversible four-state)	PASS	enrolled→disconnected (/api/storage/disconnect: unmounted, no apps) → enrolled (/api/storage/reconnect: remounted /dev/sdb1, disconnected mark cleared)
Decommission → re-enroll (non-destructive)	PASS	/api/storage/decommission mode=anyway (type-to-confirm felhom-usb): unmounted, decommissioned=True soft-marker, fs intact — uuid da9e7089 preserved, NOT formatted. /api/storage/reconnect: reenrolled=true, MOUNTED, marker cleared, label preserved
Destructive format/wipe on loopback scratch	BLOCKED (documented)	created /dev/loop0 (truncate 2G + losetup); /api/storage/init → agent refusing to operate on non-whitelisted block device "/dev/loop0" — a deliberate device-whitelist safety control. So the clean scratch target the runbook prescribed is unavailable. Per §0.3's escape clause (no clean scratch target → skip), the destructive format was not run on the live felhom-usb either. Loopback torn down.
Migrate (data)	DEFERRED	no safe scratch target (loopback blocked); a live migrate would move a real app's data — not run unattended. internal/stacks/migrate.go has unit coverage.

Findings:

1. Agent device whitelist refuses non-whitelisted block devices (loopback) — good security, but means a loopback can't serve as the destructive-disk scratch target the runbook assumed. Destructive format/migrate now need either a real spare disk or a supervised run.
2. FileBrowser sync logs mkdir …/userdata: permission denied when ensuring the skeleton on a just-unmounted/decommissioned path (recurs on disconnect + decommission). Cosmetic (the drive is intentionally unmounted), but noisy — could gate the skeleton-ensure on "is the path a live mountpoint."

Phase 3 — Break-and-recover chaos (C1–C8) — PASS (7/8 run, C4 skipped; 2 findings)

#	Break → Detect → Recover → Verify	Result	Evidence
C1	docker kill navidrome → controller detects → restart via UI → healthy	PASS	killed → Exited(137); controller state stopped; POST /api/stacks/navidrome/restart → running/healthy, state running
C2	docker kill felhom-controller → (expected bootstrap self-heal) → recover → verify	PASS (w/ 🔴 finding)	killed → stayed Exited 48s, NO auto-recovery (docker unless-stopped didn't fire; bootstrap.service is oneshot/exited). Data plane unaffected (apps+cloudflared+traefik up). Recovered via systemctl restart felhom-controller-bootstrap.service → running/healthy; UI 200; 26 containers/0 unhealthy. Finding: a crashed controller has NO automatic recovery.
C3	stop felhom-agent → controller degrades gracefully → restart agent → immediate success	PASS	agent down → /api/disks fast ~98ms clean error ("connection refused"), no hang/crash; agent restarted → /api/disks OK on the first try (4/4), no stale-conn errors; idle sockets bounded at 1 — the v0.74.0 client's restarted-agent drain works

| C4 | fill a drive near-full → disk warning → clear | SKIPPED(impractical) | drives are 118 GB / 916 GB and /var/lib/docker 252 GB; reaching the 90% warning / the deploy gate's max(5GB,10%)≈25 GB-free needs 100+ GB of fill — unsafe/slow unattended. Gate logic GetDockerVolumeHeadroom (hard-refuse deploy below the reserve) is present; monitoring reports per-drive usage. | | C5 | rename navidrome/app.yaml away → rescan → restore | PASS | rescan after removal: controller stays running/healthy (no crash), container keeps running (data plane), card → state=running deployed=false (graceful orphan). Restored + rescan → state=running deployed=true | | C6 | disconnect felhom-flash (9 apps) → recover | PASS (recovered) + finding | disconnect stopped all 9 apps + unmounted (UNMOUNTED, komga stopped); reconnect → felhom-flash remounted /dev/sdc1, all 9 restarted → 26 running, 0 unhealthy after start_period. Finding: during the drive-absent window the controller's userdata-belt/recreate tried to bring drive-backed apps up while the mount was gone → mkdir …/userdata: permission denied + transient Created states (rather than a clean held "missing-storage" stop). Reconnect cleared it; but the disconnect→recreate interaction is noisy — worth a guard ("don't recreate/ensure-userdata when the drive is absent"). |

| C7 | kill controller MID-quiesce → Recover resumes | PASS | triggered whole-guest backup → quiesce marker written upfront (active:true, 15 stopped_stacks) → docker kill felhom-controller mid-quiesce → marker persisted across the crash; controller restart logged [quiesce] crash recovery: a quiesce was in progress (15 stack(s) stopped) — restarting them → 26 running, 0 unhealthy, marker cleared, no stack stranded. Crash-safety contract proven. Bonus: a concurrent trigger was rejected mentés már folyamatban van (single-flight works). | | C8 | guest reboot during a backup op | PASS | triggered DB-dump + pct reboot 9201 mid-op → recovered to 26 containers/0 unhealthy; drives remounted by UUID (felhom-flash /dev/sdc1, felhom-usb /dev/sdb1, ext4); cloudflared+controller healthy; romm DB intact (17 tables — no corruption); UI 200 |

🔴 Finding (C1/C2) — docker restart policy not honored: containers with restart: unless-stopped do not auto-restart after docker kill (navidrome RestartCount=0 over two trials; dockerd logged stopping restart-manager; Docker 29.5.3, Live Restore off). The controller is not stopping them (no reconcile-stop in logs/source). Impact: no docker-level self-heal for crashed app or controller containers — the controller detects the stopped state and a UI/operator restart recovers apps, but a crashed controller itself has no automatic recovery (oneshot bootstrap + dead restart policy). Needs supervised investigation (docker daemon restart-manager behaviour) + a controller-recovery watchdog (e.g. bootstrap as a Restart=always monitor, or a systemd unit watching the container).

Phase 4 — Backup/restore edge cases — pending

Phase 5 — Host reboot (optional) — pending

---

Feature matrix — filled at end

Cleanup confirmation — filled at end

Deferred / needs supervised — filled at end