From 149afb104d17cadb7b4e95cc6ad69015c8c4fc84 Mon Sep 17 00:00:00 2001 From: codex Date: Mon, 6 Jul 2026 18:28:27 +0200 Subject: [PATCH] docs(CUST-WP-0054): record Phase 1 empty stack evidence on railiance01 Phase 1 deploy complete: CNPG state-hub-db, Alembic at head, empty hub healthy on ClusterIP 10.43.68.154. Documents gitea image workaround and Forgejo rebuild prerequisite before Phase 2. --- docs/state-hub-railiance01-cutover-phase1-2026-07-06.md | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/docs/state-hub-railiance01-cutover-phase1-2026-07-06.md b/docs/state-hub-railiance01-cutover-phase1-2026-07-06.md index 94f5a52..327df8f 100644 --- a/docs/state-hub-railiance01-cutover-phase1-2026-07-06.md +++ b/docs/state-hub-railiance01-cutover-phase1-2026-07-06.md @@ -73,9 +73,10 @@ Post-Phase-2 restore must match coulombcore preflight counts exactly (see Helm values file targets Forgejo (`main-f9f0091`) for post-fix cutover. Rebuild Forgejo image before Phase 2 so production and CI share one registry. -**Root cause (suspected):** CI DinD build completed without failing on empty -requirements install; investigate `Dockerfile.ci` generation and add build-time -`import asyncpg` gate in workflow. +**Root cause:** DinD layer cache reused an empty `requirements.txt` layer from an +earlier build; CI reported success at `f9f0091` but venv was empty. Fixes pushed +to Forgejo `state-hub` @ `96b7306` (Dockerfile gates + `DEPS_LOCK_ID` cache bust); +image workflow still failing on runner — rebuild tracked before Phase 2. ## Phase 2 prerequisites