Repository navigation
Deploy: gate production on a CI-built/tested image + readiness health-check + post-deploy smoke test #266
Description
Deployment-readiness gap (#244), BLOCKER — flagged by 2 of 3 reviewers.
Problem
Nothing gates production on a tested artifact, and the deployment artifact is first
exercised at deploy time:
- CI never builds the Docker image.
.github/workflows/ci.ymlbuilds/tests
.NET and web source but never runs the multi-stageDockerfile(npm ci, the
web/dist→wwwrootcopy, the sourcemap move). A Dockerfile break ships
undetected. - No container-level readiness wired. The app exposes
/health/live(process)
and/health/ready(DB reachable + migrations applied)
(Program.cs:733-737), but
the composeappservice has nohealthcheckand the Dockerfile no
HEALTHCHECK, so an instance whose/health/readyis 503 (DB down, migrations
pending) still looks up.
Fix
App-side / portable (this repo):
- Add a
docker buildCI job: build the image, boot it against a throwaway
Postgres, hit/health/ready. Identify the artifact by immutable commit
SHA/digest. - Add a
HEALTHCHECKto the Dockerfile (and a composehealthcheck) that probes
/health/ready.
Provider-side (deployment repo):
- The host's deploy manifest declares the readiness path (
/health/ready) with a
timeout covering the boot migration. - Gate promotion on CI (required checks / "wait for CI") so a red run can't deploy.
- Post-deploy smoke test through the production URL:
/health/live,
/health/ready, SPA loads, login, refresh cookie, one authenticated read.
Verify
Break the Dockerfile on a branch → CI fails. A 503-ready instance is not promoted.
Part of the #244 deployment-readiness audit; tracked in epic #15.
Activity
chore(docker): image hardening — non-root user, pinned digests, Trivy…
mforce commentedon Jul 30, 2026
Re-scoping. The deploy-side of this issue moved to a separate deployment/ops repo (deploy manifest — migrate-before-serve, readiness gate, CI-gated immutable artifact — plus a post-deploy smoke test).
Of the app-side/portable half:
HEALTHCHECK— still open: noHEALTHCHECKinsrc/Cluckwork.Api/Dockerfileand nohealthcheckon the composeappservice (onlydbhas one). Add both, probing/health/ready.Keeping this open, now scoped to just that HEALTHCHECK task (a small app-side change); everything else here is either done (#294) or relocated to that deployment repo.
feat(api): container HEALTHCHECK via in-process `healthcheck` verb (#266
mforce commentedon Jul 30, 2026
App-side landed in PR #304 (merged 2026-07-30): the runtime image + compose
appservice carry aHEALTHCHECK(an in-processhealthcheckCLI verb probing /health/ready), a/health/{**rest}→ 404 guard so the SPA fallback can't fake a green health check, and a CIimage-job smoke that boots the built image against a throwaway Postgres and asserts /health/ready=200 + Docker healthhealthy+ verb exit 0.The remaining scope is host/deploy config, tracked in a separate deployment/ops repo:
Closing here as a dup of that deploy-side work (same pattern as #260/#263); go-live tracking lives in that separate deployment repo.
docs(agents): drop the private deploy repo's name from the host-agnos…
feat(web): serve a per-request CSP nonce so Emotion's styles apply un…