Backups

Backups protect against operator error, corruption, ransomware, and regional failure. Define RPO (how much data you can lose) and RTO (how fast you must restore). Mechanisms: logical dumps, storage snapshots, continuous archiving (WAL/binlog), and cross-region replication—not interchangeable.

| Type | Pros | Risk | |------|------|------| | Logical dump | Portable, selective restore | Slow at huge scale | | Snapshot | Fast capture | Crash consistency without quiesce | | PITR (WAL) | Point-in-time | Replay chain must be intact |

Restore drills prove backups work; untested backups are wishful thinking. Encrypt at rest and in transit; restrict who can delete backup objects.

On interviews: state RPO/RTO for a payment service vs internal wiki; describe last successful restore test; mention immutability and off-site copies.

Common pitfalls: backup without restore runbook; snapshots only in same region; no retention policy; secrets in dump files; assuming replica equals backup.

The trade-off is storage cost and operational overhead versus survivable recovery when the primary is gone or corrupted.

Checklist:

• Define RPO and RTO for the service.
• Name backup type: logical, snapshot, PITR.
• Schedule and audit restore drills.
• Encrypt and tier retention (daily/weekly/monthly).
• Separate backup credentials from production admin.