mirror of
https://github.com/shankar0123/certctl.git
synced 2026-06-07 16:41:36 +00:00
shankar0123
3b96b3561c
The 25194251740 CI run failed with "container certctl-test-server is unhealthy" but the GitHub Actions log doesn't include the server's stdout/stderr — compose only reports the dependency-chain symptom. Without the server's actual log output we can't tell whether the unhealthy state was caused by a DB migration crash, port bind failure, entrypoint stall, OOM kill, or healthcheck race. Add an `if: failure()` step right before teardown that dumps: - `docker compose ps -a` (every container's exit status) - last 200 lines from certctl-test-server - all of tls-init (one-shot, short) - last 100 lines from postgres + stepca + agent - last 50 lines from pebble This is a permanent debuggability improvement, not a band-aid: the matrix-collapse (Phase 5) brings up ~18 containers concurrently where pre-collapse the per-vendor matrix brought up ~7. Future transient failures will be much faster to diagnose with logs in the CI output. Once we know the actual root cause from this dump, we fix it for real. Placed AFTER skip-count enforcement (so failures in either step trigger it) and BEFORE teardown (which is `if: always()` and would otherwise nuke the containers before we could log them).