[SOLVED] ver 2 cannot see ver 1 backups

[M_Allen]

Was running Proxmox 6x and PBS 1x. Daily backups, garbage collections, prunes and verifies all worked well. But the July 31st deadline was looming.

Updated Proxmox to 7x and PBS to 2x (most recent for both). Ran a previously defined daily backup with no problems.

The problem I do have is that PBS 2.2-3 cannot see the backups made with PBS version 1x.

No namespaces are defined.

The new backup directories are in the same directory as the old backups. (So all of VM 103's backup directories are in the 103 directory.) All of the old directories are still there and all directories (new and old) are owned by the same user, backup:backup.

The Dashboard knows there are more backups in the datastore. They just aren't shown in the datastore's Content display.

It would seem that this is a simple thing to fix but I've not been able to come up with the answer nor find anything in my many internet searches.

Any ideas?

Thanks,

Mike

 

[fabian]

did you change anything configuration wise on either the PVE or PBS side? nothing changed about the datastore layout except for the recent addition of namespaces (which are opt-in and fully compatible otherwise).. can you try doing a force-reload of the PBS browser tab to rule out any caching issues in the frontend? does PVE also only show a subset of backups, or all of them?

 

[M_Allen]

Nothing was changed.

Just updated PVE and then PBS and then ran another backup. That's when the I noticed the previous backups were not showing.

PVE only shows the new backups as well.

When in PBS in the datastore area and clicking on Content the display shows the backup container/vms but the count only includes the ver 2 backups.

 

[fabian]

what do you mean when you say "The Dashboard knows there are more backups in the datastore."? the "Task Summary" panel only shows you that backup tasks have run in the last month, not that those backups still exist.. is it possible you accidentally pruned your datastore?

 

[M_Allen]

Yes, I was assuming that the Dashboard's Task Summary was smarter than it is.

Nothing has been pruned - other than the daily schedule. The previous backups still exist along side the new backups.

BUT, now that I've have more than one new backup there is one thing that sticks out; the file names of the old version backups have odd characters in the file names. A special character is displayed instead of a colon in the time part of the file name. I had thought it was just an encoding anomaly from viewing the Shell in a browser.

drwxr-xr-x 2 backup backup 0 Jul  2 00:02 2021-12-26T150001Z
drwxr-xr-x 2 backup backup 0 Jul  8 07:53 2021-12-26T15:00:01Z

The first backup directory was made by ver 1 of PBS. I made a copy of it but used ":"s in the time for the copy. The copy "does" show up in PBS and is available for restoring in PVE.

So the solution is to just rename the backup directories using colons ":" and PBS can see the old backups.
* For some reason only the ":" is encoded differently than the other characters in the directory name.
* The ":" is shown differently when viewed over a SFTP connection in Caja.
* It is shown differently both before and after replacing the ":" with the correct character.
* The correct character is also encoded differently from the other characters in the directory name.
* Dolphin just shows small boxes - both before and after the "repair."
* Care must be taken when replacing the ":"s. A test should be made to ensure the correct character is used.

Update:
* I was able to restore a file from a ver 1 backup whose directory name had be renamed.
* I chose a text file so that it would be easy to quickly tell if it was restored correctly - and it was!
Update 2:
* I had another similar system which had the same problem after upgrading.
* Prior to upgrading the ":"s were shown as ordinal character 58.
* After upgrading to Debian Bullseye the ":"s are shown as ordinal character 61498.
* It doesn't make sense that a new version of Debian would change how it represented a character.
* Instead it would seem that the datastore was updated by the upgrade and the characters were changed. But that's just a guess.

Thank you for your time and for a "Great" product!

Mike

 

[fabian]

that does sound rather odd.. is the datastore located locally? if so, on which FS? if not, over what kind of transport (NFS/..) and if applicable, FS, is it located?

 

[M_Allen]

PBS is in a PVE VM.
* Datastore is on a Synology NAS on the LAN using BTRFS file system.
* Mounted in VM using CIFS

An interesting thing is that if I SSH into the Synology the data is owned by SynologyUser:users. From the PBS shell the data is owned by backup:backup. Regardless everything works as expected.

Only the colons ":" in only the PBS directory names have a special encoding on the Synology. Special in that they don't look like ASCII characters in the shell/terminal.