avav25 Deploy to Production
Production deployment skill providing rollback procedures, health check patterns, and verification checklists. Ensures safe, repeatable production deployments.
When to Use
- Deploying application updates to production
- Planning rollback strategies before deployment
- Verifying production health after deployment
- Investigating post-deployment issues
When NOT to Use
- Deploying to staging (use
deploy-stagingskill workflow) - Making infrastructure changes (use
infra-changeskill workflow) - Planning release version and changelog (use
releaseskill workflow)
Deployment Strategies
| Strategy | When to Use | Risk | Downtime |
|---|---|---|---|
| Rolling update | Standard deploys, stateless services | Low | Zero |
| Blue-Green | Critical services, instant rollback needed | Low | Zero |
| Canary | High-risk changes, gradual validation | Medium | Zero |
| Recreate | Stateful services, breaking schema changes | High | Brief |
Rolling Update (Default)
- Gradually replace old pods with new ones
- Health checks gate the rollout — unhealthy pods stop the rollout
- Configure
maxSurgeandmaxUnavailablefor rollout speed - Monitor error rates during rollout window
Canary Deployment
- Deploy new version to small subset (5–10% traffic)
- Monitor error rate, latency, and business metrics for 5–10 minutes
- If healthy → gradually increase traffic (25% → 50% → 100%)
- If degraded → route all traffic back to stable version
Blue-Green Deployment
- Deploy new version alongside current (green alongside blue)
- Run full smoke tests against green
- Switch traffic from blue to green
- Keep blue running for instant rollback (15–30 minutes)
- Decommission blue after confidence period
Pre-Deployment Checklist
- All tests pass on release branch
- Staging deployment verified
- Database migrations tested (forward and rollback)
- Feature flags configured
- Monitoring dashboards ready
- On-call team notified
- Rollback plan documented and tested
- Release notes prepared
Post-Deployment Verification
- Health endpoint returns 200
- Error rate ≤ pre-deploy baseline
- Latency P95 within SLO target
- No new error patterns in logs
- Key user journeys functional
- Database connections stable
- External integrations responding
- Alerts not firing
Integration
- Follows rules:
devops-engineerrole (infrastructure),sre-engineerrole (SLOs, monitoring) - Used by workflows:
deploy-productionskill,deploy-stagingskill - Companion resources:
rollback-procedure.md