Details
Assignee
UnassignedUnassignedReporter
daniel.almeidadaniel.almeidaNeeds QA
YesComponents
Priority
Medium
Details
Details
Assignee
Unassigned
UnassignedReporter
daniel.almeida
daniel.almeidaNeeds QA
Yes
Components
Priority
Smart Checklist
Smart Checklist
Smart Checklist
Created last week
Updated last week
This feature request originated from the behavior described in https://perconadev.atlassian.net/browse/K8SPSMDB-1314 .
This feature request would enable PBM to monitor storage changes and automatically update metadata in the PBM system collection. This could be potentially part of a broader active-active PBM mode for shared storage.
The scenario is as follows:
Two clusters:
Cluster A with storage X and PITR enabled (PBM uploads oplogs).
Cluster B with storage X but PITR disabled.
When running
pbm-agent status
on Cluster B, the output shows:PITR incremental backup: ======================== Status [OFF] Currently running: ================== (none) Backups: ======== S3 us-east-1 http://minio-service:9000//operator-testing Snapshots: 2025-03-24T12:41:26Z 5.85MB <physical> [restore_to_time: 2025-03-24T12:41:29Z] PITR chunks [9.17MB]: 2025-03-24T12:41:30Z - 2025-03-24T17:26:11Z
In this case, Cluster B can see backups and PITR chunks from Cluster A and successfully restore them up to 2025-03-24T17:26:11Z.
Issue
Cluster A continues uploading logs and creating new backups, but Cluster B does not sync this new data. As a result, it cannot restore beyond the initially visible PITR window.
Temporary Workaround
Run
pbm config --force-resync
manually on Cluster B before restoring to fetch the latest metadata. Then, use the updated data to create the restore object.The workaround works, however customers are looking for an automated process to update this information instead of having to run
force-resync
every time. Having to follow this process during an outage or emergency situation/high priority issue where they need to restore data quickly might be an issue as the force resync can take some time to run and this could impact their recovery SLA.