bug(compose): up -d --wait treats create-buckets exit(0) as failure #510

New Issue

marcel · 2026-05-11T16:32:12+02:00

marcel commented

2026-05-11 16:32:12 +02:00

Summary

docker compose -f docker-compose.prod.yml ... up -d --wait exits non-zero even when every service is healthy, because the one-shot create-buckets service exits 0 and --wait interprets that as "not running, fail".

Reproduction (from run 1496 log)

 Container archiv-staging-db-1 Healthy
 Container archiv-staging-ocr-service-1 Healthy
 Container archiv-staging-mailpit-1 Healthy
 Container archiv-staging-minio-1 Healthy
 Container archiv-staging-backend-1 Healthy
container archiv-staging-create-buckets-1 exited (0)
  ❌  Failure - Main Deploy staging
exitcode '1': failure

Stack functionally up. Smoke step never runs because up -d --wait fails first.

Fix

create-buckets: add restart: "no" to declare one-shot intent.
backend.depends_on: add create-buckets: condition: service_completed_successfully.

With both in place, Docker Compose v2.20+ understands create-buckets is a one-shot that must complete successfully, and --wait treats exited(0) as the target state. The backend correctly gates startup behind bucket bootstrap (closes a latent race: backend currently starts in parallel with create-buckets even though it needs the bucket + policy to exist).

Discovered

While running nightly.yml against staging after #506/#508 unblocked the prior layers. The runtime stack is now healthy (verified via direct curl against staging.raddatz.cloud); only the workflow's own wait gate fails.

## Summary `docker compose -f docker-compose.prod.yml ... up -d --wait` exits non-zero even when every service is healthy, because the one-shot `create-buckets` service exits 0 and `--wait` interprets that as "not running, fail". ## Reproduction (from run 1496 log) ``` Container archiv-staging-db-1 Healthy Container archiv-staging-ocr-service-1 Healthy Container archiv-staging-mailpit-1 Healthy Container archiv-staging-minio-1 Healthy Container archiv-staging-backend-1 Healthy container archiv-staging-create-buckets-1 exited (0) ❌ Failure - Main Deploy staging exitcode '1': failure ``` Stack functionally up. Smoke step never runs because `up -d --wait` fails first. ## Fix 1. `create-buckets`: add `restart: "no"` to declare one-shot intent. 2. `backend.depends_on`: add `create-buckets: condition: service_completed_successfully`. With both in place, Docker Compose v2.20+ understands `create-buckets` is a one-shot that must complete successfully, and `--wait` treats `exited(0)` as the target state. The backend correctly gates startup behind bucket bootstrap (closes a latent race: backend currently starts in parallel with create-buckets even though it needs the bucket + policy to exist). ## Discovered While running `nightly.yml` against staging after #506/#508 unblocked the prior layers. The runtime stack is now healthy (verified via direct curl against staging.raddatz.cloud); only the workflow's own wait gate fails.

marcel referenced this issue from a commit

2026-05-11 16:33:36 +02:00

fix(compose): mark create-buckets as one-shot for up --wait

marcel referenced a pull request that will close this issue

2026-05-11 16:33:47 +02:00

fix(compose): mark create-buckets as one-shot for `up --wait` (#510) #511

marcel referenced this issue

2026-05-11 16:34:15 +02:00

bug(caddy): `respond @actuator 404` swallowed by catch-all `handle`; /actuator/health returns 302 #512

marcel referenced this issue

2026-05-11 16:56:19 +02:00

fix(compose): mark create-buckets as one-shot for `up --wait` (#510) #511