Database Migration Rollback Plan Builder

Generate a safe database migration plan with rollout steps, rollback procedures, validation checks, and communication notes for production changes.

Prompt Template

You are a senior platform engineer responsible for production database changes. Build a migration and rollback runbook for the following change.

Database: [PostgreSQL / MySQL / etc.]
Application stack: [framework and deployment setup]
Current schema/context: [relevant tables, traffic patterns, scale]
Planned change: [column add, backfill, rename, table split, index change, etc.]
Risk level: [low / medium / high]
Downtime tolerance: [none / short maintenance window / flexible]
Dependencies: [application deploys, feature flags, ETL jobs, replicas, analytics]

Provide:
1. **Pre-migration checklist** — backups, capacity checks, communication, dry runs
2. **Migration strategy** — exact rollout order with reasoning
3. **Validation steps** — what to verify before, during, and after deploy
4. **Rollback plan** — exact triggers, steps, and decision points
5. **Monitoring plan** — errors, latency, locks, replication lag, data drift
6. **Communication template** — update for engineers and stakeholders
7. **Post-mortem prompts** — questions to answer if anything goes wrong

Assume the audience is an engineering team running this in production. Prefer safe phased rollouts, backwards-compatible changes, and feature-flagged cutovers where appropriate.

Example Output

Migration Strategy

1. Add nullable `status_v2` column without changing reads

2. Deploy app version that dual-writes to `status` and `status_v2`

3. Backfill historical rows in batches of 5,000 with lock monitoring

4. Validate parity on 100% of rows

5. Flip reads behind feature flag for 5% of traffic, then 50%, then 100%

6. Remove dual-write only after parity holds for 24 hours

Rollback Trigger

If write latency increases by 30% for 10 minutes or parity drops below 99.9%, disable the read flag and revert app deploy before touching schema.

Monitoring

- replication lag

- deadlocks

- slow query logs

- row parity dashboard

- error rate on write path

Tips for Best Results

  • 💡Include whether the migration is additive, destructive, or a rename because the safest rollout pattern changes a lot by type.
  • 💡Mention your deployment topology, a monolith, worker fleet, and analytics jobs often need different sequencing.
  • 💡Ask the AI to produce SQL snippets only after you have a sound operational plan.
  • 💡Always request explicit rollback triggers, not just rollback steps.