Database Backups and Disaster Recovery: A Production Checklist

TL;DR — A backup that hasn't been tested is not a backup. Automate daily backups with 30-day retention, enable point-in-time recovery, test restores monthly, and keep at least one copy in a different region. On RaidFrame, all of this is automatic for managed databases.

The backup checklist

Backup types

Full backups (pg_dump)

A complete copy of your database at a point in time.

# Manual backup
rf db backup main --name before-migration
 
# List backups
rf db backups main

BACKUPS — main (PostgreSQL 16)
──────────────────────────────
NAME               SIZE     CREATED              STATUS
before-migration   2.3 GB   2026-03-16 14:00     ✓ complete
auto-2026-03-16    2.3 GB   2026-03-16 03:00     ✓ complete
auto-2026-03-15    2.2 GB   2026-03-15 03:00     ✓ complete
auto-2026-03-14    2.2 GB   2026-03-14 03:00     ✓ complete
...

Point-in-time recovery (PITR)

Restore to any second within your retention window. Uses continuous WAL (Write-Ahead Log) archiving.

rf db restore main --to "2026-03-16T14:30:00Z"

Restoring main to 2026-03-16T14:30:00Z...
  WAL replay: ████████████████████ complete
  Consistency check: ✓
 
✓ Restored to 2026-03-16T14:30:00Z
  Services restarted with new connection

Use this when:

• Someone accidentally deleted rows (DELETE FROM users WHERE... without a WHERE clause)
• A migration corrupted data
• You need to recover to a specific moment before an incident

Continuous replication

Real-time streaming to a standby database in another region:

rf db replicas add main --region eu-west-1

If the primary fails, the replica is promoted automatically. Recovery time: seconds, not hours.

Disaster recovery on RaidFrame

What's automatic

Test your backups

A backup you've never restored is a hope, not a backup.

# Restore to a test database
rf db restore main --to "2026-03-15T12:00:00Z" --target test-restore
 
# Verify data integrity
rf db query test-restore "SELECT count(*) FROM users"
rf db query test-restore "SELECT count(*) FROM orders"
 
# Clean up
rf db delete test-restore

Schedule monthly restore tests:

rf cron add "0 4 1 * *" "rf db restore main --to now-1h --target monthly-test && rf db delete monthly-test" --name backup-test

Define your recovery objectives

For most SaaS apps:

• RPO: < 1 hour (use PITR on Pro plan)
• RTO: < 5 minutes (use read replicas with auto-failover)

Self-managed backup patterns

If you're running your own PostgreSQL (on a VPS or container), here's the minimum setup:

Automated pg_dump

#!/bin/bash
# backup.sh — run via cron: 0 3 * * *
TIMESTAMP=$(date +%Y%m%d_%H%M%S)
BACKUP_FILE="/backups/mydb_${TIMESTAMP}.sql.gz"
 
pg_dump -h localhost -U postgres mydb | gzip > "$BACKUP_FILE"
 
# Upload to remote storage
aws s3 cp "$BACKUP_FILE" s3://my-backups/postgres/
 
# Delete local backups older than 7 days
find /backups -name "*.sql.gz" -mtime +7 -delete

WAL archiving for PITR

# postgresql.conf
wal_level = replica
archive_mode = on
archive_command = 'aws s3 cp %p s3://my-backups/wal/%f'

Monitor backup success

Alert if a backup hasn't completed in 24 hours. Missing backups are silent failures — you won't know until you need to restore.

Recovery procedures

Scenario: Accidental data deletion

# 1. Identify when the deletion happened (check logs)
rf logs --service api --search "DELETE" --since 2h
 
# 2. Restore to point before deletion
rf db restore main --to "2026-03-16T14:25:00Z"
 
# 3. Verify data is intact
rf db query main "SELECT count(*) FROM users"

Scenario: Corrupted migration

# 1. Rollback the deployment
rf deployments rollback
 
# 2. Restore database to pre-migration state
rf db restore main --backup before-migration
 
# 3. Fix the migration and try again

Scenario: Complete region failure

# With read replicas, automatic failover happens in seconds.
# If no replica exists:
 
# 1. Restore from latest backup in another region
rf db restore main --region eu-west-1
 
# 2. Update services to use new database
rf deploy --region eu-west-1

FAQ

How often should I back up?

Daily automated backups minimum. For critical data, enable PITR (continuous WAL archiving) so you can recover to any point in time.

Where should backups be stored?

In a different region from your primary database. On RaidFrame Enterprise, cross-region backup copies are automatic.

How long should I keep backups?

30 days minimum. 90 days for compliance (SOC 2, HIPAA). Some regulations require longer — check your industry requirements.

Should I back up Redis?

Only if Redis holds persistent data (not just cache). If Redis is a cache, it can be rebuilt from the primary database. If it holds sessions or queues, enable AOF persistence.

How do I know my backups are working?

Monitor backup completion and test restores monthly. On RaidFrame, backup failures trigger alerts automatically.

Related reading

• Scaling PostgreSQL: 1K to 1M QPS
• Multi-Region Deployment Strategy
• Cloud Cost Optimization