---
description: Release readiness, rollout, rollback, and post-release monitoring
alwaysApply: true
---

# Release & Rollback

Every release must be deployable, observable, and reversible. A feature is not ready if the team cannot monitor it or recover from it.

For rollout tooling, synthetic monitoring, SLOs, on-call, and operational reporting categories, see **devops-tooling.mdc**.

## Release Readiness
- Confirm Definition of Done, QA evidence, security review, migration status, and unresolved risks before release.
- Identify deployment target, release owner, release window, affected users, and communication needs.
- List feature flags, configuration changes, migrations, background jobs, and third-party dependencies.
- Verify observability exists for the changed behavior: logs, metrics, traces, dashboards, and alerts when applicable.

## Rollout Strategy
- Prefer small, reversible releases over large batches.
- Use feature flags, canaries, phased rollout, or dark launch for risky user-facing or infrastructure changes.
- Database changes must be backward compatible when possible and safe across rolling deploys.
- Long-running migrations must have progress monitoring and an interruption plan.

## Rollback Plan
- Every release must state how to rollback code, configuration, feature flags, and migrations.
- If rollback is not possible, document the forward-fix plan and get Product Owner risk acceptance before release.
- Do not release when rollback depends on manual guesswork or undocumented production console changes.

## Post-Release
- Monitor agreed signals after release: errors, latency, saturation, conversion, usage, and user-facing failures.
- Define who watches the release and for how long.
- Capture incidents, regressions, and follow-up work as backlog items.