Backup verified even disk has broken superblock

[proxmoximdunklenwald]

Dear All,

i have a VM (Univention with Kopano) running nice and it is backed up with Proxmox-Backupserver.
The VM has sda for boot, sdb for data and sdc for swap.

Every Backup if verified ok. Here the backup i needed:


Last weekend i had to shut down the VM and was not able to boot it up because of a bad superblock of sdb.
When i restored the backup from the day before, the server also did not start with the same error.
When i selected the backup and opened the "File Restore" it showed me all 3 disk, but only for sda and sdc there was a "part" selectable.


Why does not the backupserver also verify the disk if it is accessible or not?

Sure i was able to recover the superblock with testdisk, but a verification-warning would help if the backed up disk is not accessible.

 

[dcsapak]

Last weekend i had to shut down the VM and was not able to boot it up because of a bad superblock of sdb.
When i restored the backup from the day before, the server also did not start with the same error.

it seems the superblock was already damaged when you backed up. the backup saves your disks as is (it cannot really know if its broken or not)
this is the best reason why one should also do restore tests

 

[proxmoximdunklenwald]

Thank you Dominik for your reply.

Perhaps a featurerequest?
With the new feature of doing a filerestore, wouldn't it be nice, only to check, the accessibility of this in in the backupverificationjob?
Then i do not need to do it myself by clicking on filerestore...

Proxmox is great

 

[dcsapak]

With the new feature of doing a filerestore, wouldn't it be nice, only to check, the accessibility of this in in the backupverificationjob?

no that does not work for a couple reasons:
* the server still cannot know what should be on the backup, so 'no partitions' might not be an error
* the file-restore feature for vms only works on pve, not on a pbs server, but the verify is done on the server

IMHO, a small script for pve that automatically queries the file-restore and checks for somethings should not be that complicated,
so it should be possible to write such a thing for your use case and put it in cron (or similar)

 

[RolandK]

maybe related: https://bugzilla.proxmox.com/show_bug.cgi?id=2874