Production Rollback
Safe rollback steps, verification, and communication checklist.
Rollback triggers
- Error rate exceeds 5% for 5 minutes
- P95 latency doubles from baseline
- Critical user flow broken
- Data integrity issue detected
Kubernetes rollback
kubectl -n production rollout history deployment/backend-api
kubectl -n production rollout undo deployment/backend-api
kubectl -n production rollout status deployment/backend-apiVercel rollback
vercel rollback [deployment-url]Or via dashboard: Deployments → select previous → Promote to Production.
Communication checklist
- Post in #engineering: what's broken, which deployment, rollback in progress.
- Page on-call SRE if P1.
- Execute rollback command.
- Verify health endpoint and error rate return to baseline.
- Update #engineering with resolution status.
- Schedule postmortem within 3 business days.