Without restart, K8s reuses cached pods even with imagePullPolicy: Always
when the manifest spec hasn't changed.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Runner now runs on oci (arm64) — docker/kubectl are local, no SSH needed
- deploy-preview.py rewritten with subprocess (no os.system, no SSH)
- deploy: build image, copy prod DB, create namespace, apply manifests
- teardown: delete namespace + image
- deploy-prod: build, push, rollout restart
- Simplified all workflow files to just call the Python script
- Deleted old hera-runner
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- .gitea/workflows/test.yml: unit tests + build on every push
- .gitea/workflows/deploy.yml: auto deploy to production on push to main
- .gitea/workflows/preview.yml: PR preview environments at pr-{id}.oil.oci.euphon.net
- Bakes production DB copy into preview image (no PVC needed)
- Auto-creates namespace + deployment + ingress with TLS
- Comments PR with preview URL
- Tears down on PR close
- scripts/setup-runner.sh: act_runner installation script
Runner: hera-runner (host mode, ubuntu-latest label)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
