Production Database Recovery Policy¶
Date: 2026-06-25 Status: Accepted for the first paid pilot. Roadmap item: EP8-T4.
Decision¶
The first paid pilot is approved with the current Terraform-managed production RDS recovery posture plus the new restore-check runbook.
| Area | Pilot decision |
|---|---|
| Automated backup retention | Keep backup_retention_period = 7 for production RDS. |
| Point-in-time recovery | Use AWS RDS automated backups/PITR within the 7-day retention window. |
| Final snapshot behavior | Keep skip_final_snapshot = false; intentional DB deletion must create a final snapshot. |
| Deletion protection | Keep AWS RDS deletion_protection = true and Terraform prevent_destroy = true. |
| Availability | Keep production RDS multi_az = true. |
| Restore procedure | Restore to a new DB instance first; never restore over the existing production instance during triage. |
Recovery Objective¶
- RPO: 24 hours for non-critical data loss; target 15 minutes for PITR when AWS automated backups are healthy.
- RTO: 4 hours for a controlled restore check; 8 hours for a production data-loss incident.
Payment, subscription, lease, handover, legal artifact, and account integrity
loss should be treated as SEV1 unless proven isolated and recoverable through a
narrower application-level correction.
Evidence¶
- Architecture runbook:
docs/runbooks/production-database-recovery.md. - Production Terraform:
terraform/aws-production/main.tf. - Architecture guard:
AWS observability Terraform > documents production database recovery objectives and restore checks.
Remaining Scope¶
This closes the backup/recovery policy gate for EP8-T4. RDS health alarm discoverability is recorded separately in RDS alarm verification.