CI - Lint & Test / lint-and-test (push) Failing after 0sDetails
act_runner creates a custom Docker network per job which doesn't
inherit the --mtu=1340 setting. Using the default bridge network
ensures job containers use the correct MTU matching pod network.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
CI - Lint & Test / lint-and-test (push) Failing after 6m0sDetails
Host network breaks DinD connection. Bridge with --mtu=1340
is confirmed working for TLS connections.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
CI - Lint & Test / lint-and-test (push) Failing after 0sDetails
DinD bridge network causes gnutls_handshake failure even with
correct MTU. Use host network mode so job containers share
the pod network stack directly.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
CI - Lint & Test / lint-and-test (push) Failing after 5m15sDetails
The TLS handshake failure was caused by MTU mismatch (pod: 1340,
DinD bridge: 1500), not certificate verification.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
CI - Lint & Test / lint-and-test (push) Failing after 5m13sDetails
Workflow-level env doesn't propagate to actions/checkout git process.
Set it in runner.envs so all job containers inherit it.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
CI - Lint & Test / lint-and-test (push) Failing after 3m28sDetails
Runner CI jobs fail with gnutls_handshake() error because
catthehacker/ubuntu container doesn't trust Gitea's private certificate.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
CI - Lint & Test / lint-and-test (push) Failing after 6m47sDetails
- actions/checkout@v4 requires node runtime
- Switch from ubuntu:22.04 to catthehacker/ubuntu:act-22.04
- This image includes node, git, and other tools needed by Actions
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
CI - Lint & Test / lint-and-test (push) Failing after 8sDetails
- actions/checkout@v4 requires node, which is missing in python:3.10-slim
- Use default ubuntu-22.04 image and install python3 via apt
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Train & Deploy / train (push) Failing after 0sDetails
Train & Deploy / build-and-deploy (push) Has been skippedDetails
CI - Lint & Test / lint-and-test (push) Failing after 18sDetails
- src/train.py: example training script
- tests/test_train.py: pytest test
- requirements.txt: pandas, scikit-learn
- Fix ci.yaml: use python:3.10-slim container
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
CI - Lint & Test / lint-and-test (push) Waiting to runDetails
- Set REGISTRY_URL to harbor.inje-private.com in train.yaml
- Add docker login step with Gitea secrets (REGISTRY_USER, REGISTRY_PASSWORD)
- Register secrets in Gitea repo via API
- Add harbor config to config.yaml
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- nodegroup: nd for infra components
- nodegroup: prometheus for monitoring
- ML workloads excluded from nodeSelector rule
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Deploy act_runner v0.3.0 with DinD sidecar in soo namespace
- Auto-register to Gitea (gitea.inje-private.com)
- Labels: ubuntu-latest, ubuntu-22.04, self-hosted
- Enable Actions on repo via API
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Install DVC v3.66.1 and initialize in project
- Configure NAS remote: /ainas/dvc-storage (192.168.0.43)
- NAS already mounted at /ainas on this host
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Register Gitea repo in ArgoCD (credentials configured)
- Create ArgoCD Application: watches manifests/ → deploys to soo ns
- SyncPolicy: automated with prune and selfHeal
- Add manifests/ directory for future K8s resources
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- config.yaml: cluster/component infrastructure config
- README.md: install guide and component status
- .gitignore: exclude .env and secrets
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Cloud User