Skip to content
Percona Operators Documentation

Rate this page
☆ ☆ ☆ ☆ ☆
Thanks for your feedback
Thank you! The feedback has been submitted.

Get free database assistance or contact our experts for personalized support.

Get help from Percona

Enable point-in-time recovery (PITR)

Point-in-time recovery rolls the cluster back to a specific date and time. The Operator restores a full backup, then applies oplog on top of it. For what PITR supports and when to use it, see Point-in-time recovery.

Before you begin

  • Configure backup storage. PBM saves oplog to that storage.
  • You must have a successful backup (full logical, physical or incremental base) to serve as the base for oplog collection. Without one, Percona Backup for MongoDB does not upload oplog. Take a backup for a new cluster and after you restore from a backup. See Make a backup.

Enable oplog collection

Set backup.pitr.enabled in deploy/cr.yaml:

backup:
  ...
  pitr:
    enabled: true

After you enable point-in-time recovery, it takes 10 minutes for the first oplog chunk to be uploaded. The default interval is 10 minutes. Change it with backup.pitr.oplogSpanMin.

Multiple storages

You can enable point-in-time recovery and take backups on a storage of your choice. PBM saves oplog only to the main storage so data stays consistent. You can then restore to a point in time from any backup on any storage.

You must have a single storage in spec.backup.storages. PBM writes oplog to the same bucket as the backup snapshot. If you define several storages and enable PITR, PBM cannot guarantee consistency, so the Operator does not allow it. You will see an error in the Operator logs.

Restore options

You can restore your database on the same cluster or on a new cluster. Choose your restore path.

  • This cluster — undo a bad write, or other change on this cluster.
  • New cluster — restore that same point in time into another Kubernetes environment. Use this for disaster recovery or when you want to leave the source cluster running.

PVC snapshot (external) backups do not support this restore.

flowchart LR
  A[Enable oplog collection] --> B[Wait until oplog is stored]
  B --> C{Where do you restore?}
  C -->|Undo a change here| D[This cluster]
  C -->|Recover in a new environment| E[New cluster]

To restore a single collection under a different name to a given time, see Point-in-time recovery with namespace remapping.

Next steps

Restore to a point in time

Last update: September 2, 2026
Created: September 2, 2026