deploy-gh-k8s

name: deploy-gh-k8s description: End-to-end deployment workflow for this repo that gates on the latest GitHub Actions run, deploys with deploy/upgrade.sh to Kubernetes, monitors kubectl resources, troubleshoots failures (logs, code, IaC), and performs health checks. Use when the user asks to deploy/release/go-live, says "进行部署", "发布", "上线", or requests a Kubernetes deployment/check.

Deploy GH K8s

Overview

Execute a safe, gated deployment: verify the latest GitHub Actions workflow run is successful, then deploy to Kubernetes with deploy/upgrade.sh, monitor cluster resources, troubleshoot failures, and finish with health checks.

This skill assumes the project follows the project-bootstrap conventions:

• deploy/deploy.sh applies base manifests.
• deploy/upgrade.sh restarts deployments and waits for rollout.
• k8s/base/ contains the baseline resources.

Workflow

1. Identify the latest GitHub Actions workflow run.

• Use gh run list -L 1 in the repo.
• If a specific branch/PR is mentioned, filter to that branch or PR.

2. If the latest run failed, inspect and fix.

• Use gh run view <run_id> --json ... and gh run view <run_id> --log.
• Summarize the failure, implement fixes (code or IaC), and report back.
• Ask the user to review and re-submit (do not deploy yet).

3. If the latest run succeeded, deploy to Kubernetes.

• Run deploy/upgrade.sh from repo root.
• Monitor resources during rollout:
  • kubectl get pods -A
  • kubectl get deploy -A
  • kubectl get sts -A
  • kubectl get events -A --sort-by=.lastTimestamp

4. If deployment errors appear, troubleshoot.

• Identify failing resources and pull logs:
  • kubectl describe <resource> <name> -n <ns>
  • kubectl logs <pod> -n <ns> --tail=200
• Fix issues in code or IaC as needed, then report back.
• Ask the user to review and re-submit (do not claim success).

5. If deployment succeeds, run health checks.

• Use repo-appropriate health checks (service endpoints, readiness probes, or scripted checks if present).
• Report completion and any follow-up observations.

Notes

• Prefer actionable, minimal output in user updates: status, key failure snippet, and next required action.
• Do not attempt external CI providers; only report their URLs if encountered.