How to restore a local backup

This is a guide on how to restore a locally made backup.

To restore a backup that was made from a different cluster, (i.e. cluster migration via restore), see How to migrate cluster using backups.

Prerequisites

• Deployments have been scaled-down to a single PostgreSQL unit (scale it up after the backup is restored)
• Access to S3 storage
• Configured settings for S3 storage
• Existing backups in your S3-storage
• Point-in-time recovery requires the following PostgreSQL K8s charm revisions:
  • 382+ for arm64
  • 381+ for amd64

Summary

• List backups
• Point-in-time recovery
• Restore backup

List backups

To view the available backups to restore, use the command list-backups:

juju run postgresql-k8s/leader list-backups

This should show your available backups like in the sample output below:

list-backups: |-
  Storage bucket name: canonical-postgres
  Backups base path: /test/backup/

  backup-id            | action             | ... | timeline
  ---------------------------------------------------------------------------
  2024-07-22T13:11:56Z | full backup        | ... | 1
  2024-07-22T14:12:45Z | incremental backup | ... | 1
  2024-07-22T15:34:24Z | restore            | ... | 2
  2024-07-22T16:26:48Z | incremental backup | ... | 2
  2024-07-22T17:17:59Z | full               | ... | 2
  2024-07-22T18:05:32Z | restore            | ... | 3

Below is a complete list of parameters shown for each backup/restore operation:

• backup-id: unique identifier of the backup.
• action: indicates the action performed by the user through one of the charm action; can be any of full backup, incremental backup, differential backup or restore.
• status: either finished (successfully) or failed.
• reference-backup-id
• LSN start/stop: a database specific number (or timestamp) to identify its state.
• start-time: records start of the backup operation.
• finish-time: records end of the backup operation.
• backup-path: path of the backup related files in the S3 repository.
• timeline: number which identifies different branches in the database transactions history; every time a restore or PITR is made, this number is incremented by 1.

Point-in-time recovery

Point-in-time recovery (PITR) is a PostgreSQL feature that enables restorations to the database state at specific points in time.

After performing a PITR in a PostgreSQL cluster, a new timeline is created to track from the point to where the database was restored. They can be tracked via the timeline parameter in the list-backups output.

Restore backup

To restore a backup from that list, run the restore command and pass the parameter corresponding to the backup type.

When the user needs to restore a specific backup that was made, they can use the backup-id that is listed in the list-backups output.

juju run postgresql-k8s/leader restore backup-id=YYYY-MM-DDTHH:MM:SSZ

However, if the user needs to restore to a specific point in time between different backups (e.g. to restore only specific transactions made between those backups), they can use the restore-to-time parameter to pass a timestamp related to the moment they want to restore.

juju run postgresql-k8s/leader restore restore-to-time="YYYY-MM-DDTHH:MM:SSZ"

Your restore will then be in progress.

It’s also possible to restore to the latest point from a specific timeline by passing the ID of a backup taken on that timeline and restore-to-time=latest when requesting a restore:

juju run postgresql-k8s/leader restore restore-to-time=latest