PITR

What is PITR (Point-in-Time Recovery)?

PITR (Point-in-Time Recovery) is a database backup and recovery technique commonly used in Relational Database Management Systems (RDBMS). It allows for the recovery of data changes to a specific point in time, restoring the database to a state prior to that point. In PITR, the database system regularly creates full backups and logs all transactions thereafter, including insert, update, and delete operations. During recovery, the system first restores the most recent full backup, and then applies the transaction logs recorded after the backup, bringing the database back to the desired state.

KubeBlocks supports PITR for databases such as MySQL and PostgreSQL. This documentation takes PostgreSQL PITR as an example.

How to perform PITR?

1. View the timestamps to which the cluster can be restored.

   # 1. Get all backup objects for the current cluster
   kubectl get backup -l app.kubernetes.io/instance=pg-cluster
   
   # 2. Get the backup time range for Continuous Backup
   kubectl get backup -l app.kubernetes.io/instance=pg-cluster -l dataprotection.kubeblocks.io/backup-type=Continuous -oyaml
   ...
   status:
       timeRange:
       end: "2024-05-07T10:47:14Z"
       start: "2024-05-07T10:07:45Z"
   

   It can be seen that the current backup time range is 2024-05-07T10:07:45Z ~2024-05-07T10:47:14Z. Still, a full backup is required for data restoration, and this full backup must be completed within the time range of the log backups.

   kbcli cluster describe pg-cluster
   >
   ...
   Data Protection:
   BACKUP-REPO   AUTO-BACKUP   BACKUP-SCHEDULE   BACKUP-METHOD   BACKUP-RETENTION   RECOVERABLE-TIME                                                
   minio         Enabled       */5 * * * *       archive-wal     8d                 May 07,2024 15:29:46 UTC+0800 ~ May 07,2024 15:48:47 UTC+0800
   

   RECOVERABLE-TIME represents the time range within which the cluster can be restored.

   It can be seen that the current backup time range is May 07,2024 15:29:46 UTC+0800 ~ May 07,2024 15:48:47 UTC+0800. Still, a full backup is required for data restoration, and this full backup must be completed within the time range of the log backups.

2. Restore the cluster to a specific point in time.

   apiVersion: apps.kubeblocks.io/v1alpha1
   kind: OpsRequest
   metadata:
   name:  pg-cluster-pitr
   spec:
   clusterName:  pg-cluster-pitr
   restore:
       backupName: 818aa0e0-pg-kubeblocks-cloud-n-archive-wal
       restorePointInTime: "2024-05-07T10:07:45Z"
       volumeRestorePolicy: Parallel
   type: Restore
   

   kbcli cluster restore pg-cluster-pitr --restore-to-time 'May 07,2024 15:48:47 UTC+0800' --backup <continuousBackupName>
   

3. Check the status of the new cluster.

   kubectl get cluster pg-cluster-pitr
   

   kbcli cluster list pg-cluster-pitr
   

   Once the status turns to Running, it indicates a successful operation.