Deploy Service

Deployment Workflow

Follow this procedure for deploying a service to production. This skill covers pre-flight checks, staging validation, production rollout, and post-deployment monitoring.

Prerequisites

• You must have the run_tests tool available.
• Access to the deployment pipeline (CI/CD).
• The service must be on the main branch with all tests passing.

Step 1 — Pre-flight Checks

1. Verify tests pass — use the run_tests tool with suite: "all".
   • All unit and integration tests must be green.
   • If any test fails, stop — do not proceed with deployment.
2. Check the changelog — confirm that CHANGELOG.md has been updated for this release.
3. Review open issues — ensure no critical/blocker issues are tagged against this release milestone.

Step 2 — Version Tagging

Create a semantic version tag following SemVer:

git tag -a v<X.Y.Z> -m "Release v<X.Y.Z>"
git push origin v<X.Y.Z>

Step 3 — Deploy to Staging

1. Trigger the staging deployment pipeline.
2. Wait for the deployment to complete (typically 3–5 minutes).
3. Run smoke tests against the staging environment:
   • Health check endpoint returns 200 OK.
   • Key user flows work end-to-end.
   • No error spikes in the staging log dashboard.

If staging fails

• Rollback immediately using the previous stable tag.
• Investigate the failure, fix, and restart from Step 1.

Step 4 — Deploy to Production

⚠️ Only proceed after staging sign-off.

1. Trigger the production deployment pipeline.
2. Use a canary deployment strategy:
   • Route 10% of traffic to the new version.
   • Monitor error rates for 5 minutes.
   • If error rate stays below 0.1%, proceed to full rollout.
   • If error rate exceeds 0.1%, rollback immediately.
3. Gradually increase traffic: 10% → 25% → 50% → 100%.

Step 5 — Post-Deployment Monitoring

For the first 30 minutes after full rollout, actively monitor:

• Error rates in the logging dashboard.
• Response latency (p50, p95, p99).
• CPU and memory utilisation of service pods.
• Customer-facing alerts and support channels.

If any anomaly is detected, trigger an immediate rollback and create a post-incident report.

Rollback Procedure

# Deploy the previous known-good tag
deploy --service <service-name> --tag <previous-tag> --environment production

Always notify the team in #deployments when a rollback occurs.