rollback-runbook-author

A rollback runbook should use numbered, stranger-executable steps with one action, one observable result, and one verification per step. Include explicit decision thresholds, previous-good artifact identification, bad-window data handling, comms, rehearsal, and staleness triggers.

Rollback decision criteria must define specific symptoms, thresholds, a time-box, and the decision role. This ensures responders know exactly when to initiate backout procedures versus attempting a fix forward during a deployment incident.

Layered reversal planning maps the correct rollback primitive for traffic, flags, artifacts, config, schema, and data. It documents the proper reversal order to ensure dependencies are handled safely during backout.

Yes, rollback strategy applies to migrations including expand/contract patterns. The runbook confirms bad-window data posture and documents exactly how to verify recovery after reversing schema and data changes.

A rollback strategy for feature flags must include reversal criteria, layer-by-layer reversal primitives, verification steps, and staleness triggers. It documents how to toggle flags back and verify the previous-good state is restored.

Bad-window data handling requires documenting the data posture during the failure window and specifying exactly how to verify recovery. The rollback runbook confirms what data was affected and how to reconcile it after backout.