title: Runbook — deploy to production project: engineering-team-template tags:

• runbook
• deploy
• production date: 2026-04-07

Runbook — deploy to production

Service: api-core (template) | Owner: Platform rotation | Last verified: 2026-04-07

Preconditions

• [ ] Change merged to main with green CI and required approvals.
• [ ] Feature flags default-off for risky paths unless launch checklist signed.
• [ ] Maintenance window communicated if database migration >30s lock risk.

Steps

1. Announce deploy start in #deploys with commit SHA and ticket link.
2. Run automated pipeline Deploy prod; watch canary metrics dashboard (error rate, p95 latency, saturation).
3. Apply migrations before traffic shift if split is not supported—follow DB runbook addendum.
4. Shift 10% → 50% → 100% canary if healthy; pause ≥5 minutes between stages.
5. Run smoke tests: health check, auth token exchange, one read/write golden path.

Rollback

• If canary error rate >2x baseline for 3 minutes: abort and roll back to previous artifact.
• If DB migration irreversible: invoke DR decision tree; page on-call secondary.

Post-deploy

Tag the release; post changelog to #deploys; confirm SLO burn normalizes within 30m; update the status page if users see the change. On-call: PagerDuty api-core; DBA: internal roster. Tune stage names and thresholds to your real telemetry.