Files

codex 0bebad0bae docs(CUST-WP-0054): update Phase 1 evidence with Forgejo image fix notes

Record main-d1a4fcf CI fix and DinD root cause after image rebuild.

2026-07-06 18:59:04 +02:00

3.6 KiB

Raw Blame History

T05 Phase 1 Evidence — Empty Stack on railiance01 (2026-07-06)

Workplan: CUST-WP-0054-T05
Parent runbook: docs/state-hub-railiance01-cutover-plan.md
no_secret_material_recorded: true

Checklist

Step	Action	Result	Evidence
1.1	Create `state-hub` namespace	pass	`kubectl get ns state-hub`
1.2	Deploy `state-hub-db` CNPG	pass	Cluster `databases/state-hub-db` healthy, 1 instance
1.3	Apply DB NetworkPolicies	pass	Applied from `deploy/railiance/platform/`
1.4	Create `state-hub-env` Secret	pass	`DATABASE_URL` → `state-hub-db-rw.databases.svc.cluster.local`
1.5	Helm image pin (Forgejo target)	deferred	Values updated; deploy uses gitea workaround (see § Image)
1.6	Alembic init Job	pass	`state-hub-alembic-init` Completed; head `f0a1b2c3d4e5`
1.7	Helm install + `/state/health`	pass	Pod Running; health `ok`, db `connected`

Phase 1 status: complete (empty hub healthy; no production data). Phase 2 blocked on operator approval (Phase 0.6).

Deployed resources (railiance01-k3s)

Resource	Namespace	Status
Secret `state-hub-db-credentials`	`databases`	present
CNPG Cluster `state-hub-db`	`databases`	healthy
Secret `state-hub-env`	`state-hub`	present
Helm release `state-hub` (rev 2)	`state-hub`	deployed
Job `state-hub-alembic-init`	`state-hub`	Completed
Deployment `state-hub`	`state-hub`	Running 1/1
Service `state-hub`	`state-hub`	ClusterIP `10.43.68.154:8000`

Kubeconfig: KUBECONFIG=~/.kube/config-hosteurope (workstation WSL).

API verification (port-forward `18080:8000`)

curl -s http://127.0.0.1:18080/state/health
# {"status":"ok","db":"connected"}

curl -s http://127.0.0.1:18080/state/summary
# totals: topics=0, workstreams=0, tasks=0, decisions=0

Captured: 2026-07-06T16:24:19Z.

Database verification (empty schema)

Alembic at head f0a1b2c3d4e5. Baseline gate tables — all zero rows:

Table	Count
`workplans`	0
`tasks`	0
`topics`	0
`progress_events`	0
`token_events`	0
`managed_repos`	0
`decisions`	0

Queried via kubectl exec on databases/state-hub-db-1 as postgres.

Post-Phase-2 restore must match coulombcore preflight counts exactly (see docs/state-hub-railiance01-cutover-preflight-2026-07-06.md).

Image workaround

Image	Status	Notes
`forgejo.coulomb.social/coulomb/state-hub:main-f9f0091`	broken	Empty venv (DinD `/tmp` + cache defect)
`forgejo.coulomb.social/coulomb/state-hub:main-d1a4fcf`	fixed	CI green @ `d1a4fcf`; `asyncpg` 0.31.0, 246-byte requirements
`gitea.coulomb.social/coulomb/state-hub:f2e042a`	deployed	Phase 1 Helm override; swap to Forgejo tag before Phase 2

Root cause: DinD legacy builder did not persist /tmp across RUN steps and reused stale COPY layers; CI reported success at f9f0091 with empty venv. Fixed in state-hub @ d1a4fcf (single-RUN deps install under /app).

Phase 2 prerequisites

Operator approval for freeze/final restore (Phase 0.6)
Fix and republish Forgejo state-hub image (main-d1a4fcf)
Helm upgrade railiance01 deploy to Forgejo image before Phase 2
Fresh pg_dump from coulombcore per CUST-WP-0011-T07 playbook

References

state-hub/deploy/railiance/apps/helm/state-hub-values.yaml — Forgejo image pin
docs/state-hub-railiance01-cutover-plan.md — Phase 2–5 sequence
ClusterIP 10.43.68.154 — capture for Phase 3 tunnel rewire

3.6 KiB Raw Blame History Unescape Escape