unused/deleted chunks not being recovered

[David Thistlethwaite]

Hi folks, upgraded to pbs 3.0, thanks team.

Odd situation I decided to move about 1.8TB of data from one datastore to another
Deleted the backups/namespace from inside pbs, waited a week, no disk space freed up ?
Ran GC again, no pending removals

Any way I can force pbs to rescan all the chunks and rebuild the indices, this might help ?

Ideas ?

David

 

[fiona]

Hi,
how and what did you move exactly? Did you remove all backups/namespaces from the datastore or just some? What does the log from the garbage collection say?

 

[David Thistlethwaite]

Hi,
how and what did you move exactly? Did you remove all backups/namespaces from the datastore or just some? What does the log from the garbage collection say?

Thanks for ideas Fiona, they led me to find the issue it turns out that is was a not obvious zfs snapshot taking up the space, which of course makes sense after such a large deletion, but my normal process of examining snapshots did not immediately present the issue.

Deleted all existing snapshots and space is recovered. (I use snapshot as a potential ransomware protection mechanism).

Thanks again

 

[fiona]

Thanks for ideas Fiona, they led me to find the issue it turns out that is was a not obvious zfs snapshot taking up the space, which of course makes sense after such a large deletion, but my normal process of examining snapshots did not immediately present the issue.

Deleted all existing snapshots and space is recovered. (I use snapshot as a potential ransomware protection mechanism).

Thanks again

Glad you found the culprit! FYI, there is zfs list -o space where you can see the snapshot usage for each dataset right away or zfs list -o space -t all to show each individual snapshot's usage too.