Skip to main content
Version: Next

Point-in-time recovery

Point-in-time recovery (PITR) continuously uploads an instance's transaction log — such as PostgreSQL WAL, MySQL binary logs or the MongoDB oplog — to a backup storage. Together with a full backup, this lets you restore the instance to any moment within the recovery window, not only to the moment a backup was taken. Use it to recover from accidental writes or deletes.

PITR is available only for providers whose backup class supports it. The Plugin Hub shows which providers do.

Enable PITR​

PITR builds on full backups, so the instance needs at least one backup schedule on the storage first.

  1. Open the instance and go to the Backups tab, or open the Backups step when you create the instance.
  2. Turn on Point-in-time Recovery and select the backup storage to upload the log to.
  3. Optionally, adjust the PITR Configuration. The available settings come from the backup class, for example the upload interval.
  4. Save.

With kubectl, set pitr.enabled on a storage entry of the instance:

spec:
  backup:
    storages:
      - storageRef:
          name: s3-backups
        schedules:
          - name: daily
            enabled: true
            cron: "0 2 * * *"
        pitr:
          enabled: true

Some engines can upload their log to only one storage at a time — for example MongoDB and MySQL. Others, such as PostgreSQL, can archive to every storage.

Check the recovery window​

The instance reports, for every PITR-enabled storage, the earliest and latest time you can restore to:

kubectl get instance orders-db -n demo -o jsonpath='{.status.backup}'

The window starts at the oldest full backup that has a continuous log after it. If the log has a gap — for example after PITR was turned off for a while — the window starts after the gap.

Restore to a point in time​

See Restore a backup and Create an instance from a backup.