Backup, Restore, and Point-in-Time Recovery
What the exam tests
This domain tests SLA design, Time Machine behavior, restore choices, and PITR boundaries.
Core rules
- An SLA defines snapshot frequency, retention, and log backup cadence. The Time Machine applies that SLA to a database.
- Snapshots provide recovery points; log backups provide continuous recovery between snapshots for engines that support PITR.
- PITR restores to a timestamp by combining a base snapshot with logs. The achievable window is bounded by the oldest retained snapshot and log retention.
- Restore to a new database when possible; in-place restore overwrites the source and needs stronger change control.
- A restore is a provisioning workflow: it needs compute, network, project, and storage capacity.
- Test restores prove both backup integrity and operational readiness.
Worked example
At 14:05 an operator deletes a table in MySQL. The last snapshot is 13:00, and logs are retained. The administrator chooses PITR to 14:04, restores to a new database in the recovery project, validates row counts, and then exports the table. Restoring to 14:04 rather than 14:05 avoids replaying the deletion. If log retention ended at 13:30, 14:04 would be impossible.
Traps
Do not confuse a clone with a PITR restore; a clone uses a snapshot and does not replay logs. Do not assume snapshots alone give second-level recovery. Do not ignore time zone when selecting a PITR timestamp. Do not delete the snapshot or log chain that supports the required recovery window.