commvault.com commvault.com
Question

PostgreSQL database restore FSBasedBackup

  • 29 April 2024
  • 2 replies
  • 42 views
M
Rank
Badge +3

Anyone have some insight on the following:

We have a FSBasedBackup set for multiple PostgreSql databases. Databases are quite big in size. 

There is a full backup scheduled to run once per day and log (incremental) backups hourly. Full backups are taking 1 to 3 hrs to complete.

We often execute planned "refreshes" (restore point in time) of these backups to another environment.

The idea is, if i select a restore point in time from a timeframe when the full backup was still running (full backup backup already completed successfully), Commvault does not discover any backup data during browse. If we choose to restore to any other point in time outside of the full backup, restores are working fine.

Is this working as expected? Does this means that i do not have any recovery point during the timeframe when the full was running (looks odd to me)? If so, why Commvault does not show the last available data until the point i selected?

"FSStartRestoreService::prepareForPostgreSqlRestore() - Browse response did not contain any result items. check Browse.log on MediaAgent"

Version: 11.28

2 replies

Alex Deaconu
Rank
Userlevel 1
Badge +2

Hello Mihai,

 

This sounds like a case would be needed to investigate. For right now, can you send us a few Screenshots of the steps you are following to restore, and the screen you are seeing with the error message?

 

If Commvault does not allow you to browse to a Point in Time between when the Full Backup started and when it ended, then as a quick workaround, if needed, you could do a Data Restore only, browsed from the Full Backup job itself, and then start the PostgreSQL Server with the proper PIT configuration in postgresql.conf.
https://documentation.commvault.com/2023e/expert/performing_application_free_restore_of_postgresql_server_to_disk_and_then_recovering_it.html
 
The postgresql.conf options that need to be changed are ‘recovery_target_time’ and ‘restore_command’.

 

As mentioned, this is only if you need a quick workaround. Otherwise, we can continue working on this issue.

M
Rank
Badge +3

Hi Alex,

Thanks for the workaround, but the idea is many of those restore jobs are automated and using the workaround implies additional work.

Need to understand exactly how it is working so we can adapt.

Ended up by opening a case.

Reply


Terms & ConditionsCookie settings