SDD Classification: L4-Operational | Authority: Platform Engineering | Review Cycle: Quarterly
Triage
Rollback when:- availability is degraded
- error rate spikes
- a deployment introduces incorrect behavior with no safe runtime mitigation
Application rollback (Kubernetes)
GitOps rollback
If using GitOps, rollback is usually:- Revert the deployment commit
- Allow controller reconciliation
- Verify health + telemetry
Database-aware rollback
- If migrations are backwards compatible, an app rollback is usually safe.
- If migrations are breaking, you need a downgrade plan (or restore) before rolling back the app.
Verification checklist
- Health checks pass
- Error rate returns to baseline
- Key dashboards stable
- Incident summary captured