RUNBOOK: Authentication Module Rollback

Objective: Restore the service to a stable version in less than 15 minutes in case of a broken deployment.

1. Preparation (Problem Identification)

• [ ] Observe the error (Consecutive failing health checks /health in production, rapid spikes of 5xx errors on /auth/login).
• [ ] Identify the last known working version (Commit hash).

2. Rollback Execution (Kubernetes / Docker)

Via Kubernetes

# Identify deployment history
kubectl rollout history deployment/auth-service --namespace=production

# Restore previous revision (n-1)
kubectl rollout undo deployment/auth-service --namespace=production

Via Docker Compose

# Modify the YAML image tag (in docker-compose.yml) to the previous version
docker-compose up -d --build

3. Verifications (Post-rollback)

• [ ] Verify that K8s Pods (or Docker containers) restart and reach 'Running' status.
• [ ] Manually call the health check endpoint (curl -sf https://api.example.com/health).
• [ ] Confirm the absence of errors via logs.
• [ ] Validate with a basic login smoke test that authentication has returned to normal.

4. Investigation

• [ ] Extract the logs that led to the crash (before cleanup).
• [ ] Create a high-priority incident ticket to fix the "new" unstable version.