Rollback, Recovery, and Incident Decisions A rollback plan must exist before the release. Roll back application code quickly when code/config is the cause; treat database/schema incidents separately because destructive migrations may not be reversible by restoring an old image.
Investigate root Bad release Stop promotion Rollback app Verify health → → → → cause
API rollback • Railway can roll back to a previous successful deployment while restoring that deployment's image and custom variables within the platform retention window.
• Use the last known-good immutable SHA, then rerun smoke checks.
Web rollback • Promote a known-good previous Vercel deployment or redeploy the known-good commit according to team policy.
• Keep production environment variables version/change-controlled operationally even though they live outside Git.
Database incident • Do not assume an application rollback fixes a destructive migration.
• Use backup/PITR or a forward-fix migration according to the incident runbook.
• This is why expand-contract and tested migrations are part of deployment design. Rollback is a capability, not a strategy. Recovery decisions depend on whether the failure is code, configuration, data, migration, dependency, or infrastructure.
Sources: Railway deployment rollback behavior; Vercel immutable deployment model; Supabase backup/PITR production guidance.
Deployment is the controlled movement of a versioned artifact into a configured runtime with evidence of health and a credible rollback path.
Restate the goal of “Rollback Recovery and Incident Decisions” as one observable behavior. Write the success signal before touching code.
Trace the boundary map: versioned artifact → production config → runtime → health/smoke test → rollback path. Mark the one boundary this lesson changes.
Inspect the current repository state and record the exact files, route, table, or runtime involved. Avoid creating an abstraction before you can name the existing owner.
Implement the smallest vertical slice. Add one coherent change, run the narrowest verification command, and read the output before continuing.
Connect the slice to TaskFlow’s phase artifact: Staging and production deployment runbooks with release identity, environment matrix, smoke tests, and rollback procedure.. Verify behavior from the outside as a user or API client would.
Create one controlled failure that is safe to reproduce. Predict the symptom first, capture evidence, locate the earliest broken boundary, then fix only that cause.
Close the reference and rebuild or explain the core change from memory. Record what you would monitor or test before trusting the same change in production.
1. Build an immutable artifact.
2. Apply production configuration/secrets.
3. Run compatible migrations.
4. Start the new version.
5. Check readiness and smoke tests.
6. Observe errors/latency.
7. Promote or execute rollback.Do not copy this as a finished answer. Predict what each boundary does, type the smallest useful slice, run it, then compare the result with your prediction.
Separate platform deploy status from application readiness; run smoke tests against real routes and dependencies.
Use expand/contract migrations and keep rollback compatibility explicit before release.
Without copying the final source, create a second small scenario that applies “Rollback Recovery and Incident Decisions” to a different TaskFlow resource or input. Write the acceptance criteria first, implement it, and provide the evidence that each criterion is satisfied.
Configure custom domains/TLS and the exact CORS/redirect origins for production.
Use the smallest change that demonstrates this lesson. Stop after it works and capture evidence.Code deploy occurs before required migration or vice versa without compatibility plan
Predict the symptom before causing the failure. Then locate the earliest boundary where reality diverges from expectation.Why does this responsibility live here, what input does it trust, what output proves success, and what would go wrong if another layer owned it?
If your explanation is vague, repeat the boundary trace before moving on.Close the reference. Reproduce the core idea from memory or explain the exact request/data flow aloud. Then check your diff or notes for accidental shortcuts that bypass validation, authorization, error handling, or verification earned in earlier phases.