What they are (brief):
- Log-based recovery: replay transaction logs (WAL/binlog) to advance a baseline.
- PITR: restore to a specific time/transaction—usually done by restoring a baseline and replaying logs up to that point.
When to use:
- Use logs + PITR for mission-critical systems needing seconds/minutes RPO.
- Snapshots or periodic backups are fine when hours/days of data loss are acceptable.
Pros / Cons (one-liners):
- Logs: very low RPO and storage-efficient, but operationally complex and adds continuous IO.
- PITR: precise time-point recovery, but multi-step and can increase RTO if many logs must be replayed.
Quick checklist (3 items):
- Define RPO/RTO per service.
- Keep regular baselines + archived logs (retention window).
- Automate and test PITR regularly.
