Date: 2026-05-13 UTC
Scope: autocoder-engine namespace, Deployment/autocoder-engine
Approval: operator approved manual recovery in Codex chat.
Gitea Actions run 703 for AutoCoder/AutoCoder-Engine blocked in deploy_manifests while waiting for Deployment/autocoder-engine to become ready. The live backend image gitea.app.lan/autocoder/autocoder-engine:sha-b55ed8746d84 was crash-looping during startup with a PostgreSQL type mismatch error:
ERROR: operator does not exist: name[] = text[]
The linked fix had been merged to origin/main at 2e1b8098966a, but no registry image existed yet for sha-2e1b8098966a. Because the workflow gates backend image build behind successful manifest rollout, the automated pipeline could not build the fixed image by itself.
origin/main commit 2e1b8098966a locally from a temporary worktree under ~/temp/autocoder-engine-rollout-703.gitea.app.lan/autocoder/autocoder-engine:sha-2e1b8098966a.Deployment/autocoder-engine container image and AUTOCODER_ENGINE_IMAGE to sha-2e1b8098966a.AUTOCODER_WORKER_IMAGE on sha-b55ed8746d84, matching the workflow path because this was not a worker-runner change.Deployment/autocoder-engine rollout readiness./healthz returned "ok" from inside the backend pod.After the engine recovered, it processed pending issue automation and advanced origin/main to b9f60c26a10b (AutoCoder: Merge issue #2392 - pat settings stuck open (#2407)). Gitea Actions run 704 then completed successfully:
deploy_manifests: successbuild_backend_image: successrollout_backend: successdeploy_cronjobs: successdeploy_frontend: skippedFinal live backend state:
Deployment/autocoder-engine image: gitea.app.lan/autocoder/autocoder-engine:sha-b9f60c26a10b
Ready replicas: 1/1
Backend pod restarts: 0
Health check: ok
Run 703 remained failed because it had already timed out on the broken image before manual recovery completed. Run 704 is the successful automatic rollout for the current main branch.