Health check on copy job on repository with immutability

[markus.wild]

Hello,

we're planning to deploy linux hardened repositories, which will be targets of backup copy jobs with GFS rules.
It is my understanding that all these GFS generations will be set to immutable during their lifetime, which is great.
Since this will use XFS based block-relinking, those GFS backups will likely be using synthetic full backups, correct?
Now, since this is XFS based and not ZFS, I'd like to also perform periodic health checks on these backups, but I don't know what I should expect to happen if actual checksum errors should be found during such a check. The backup file itself is immutable, and can't be fixed. Would B&R just throw an error in this case, or would it attempt to recreate a new instance of the broken backup file, if that is at all possible? If a backup file is detected to be broken, but is needed to perform the next synthetic backup, will things just fall apart?
Thank you for some insights

Regards,
Markus

[foggy]

Hi Markus, valid data blocks for the corrupted restore point are stored in the new incremental file together with data blocks for the new restore point.

[markus.wild]

Thanks for the quick answer! That linked article indicates, that the health check does in fact just validate the most recent chain. Would we have to explicitly run validator to periodically detect possible errors in previous backup generations? Can validator run concurrently with an active backup job, or will it block the job like the health check while it's running?

[foggy]

Would we have to explicitly run validator to periodically detect possible errors in previous backup generations?

Depends on whether you want to be sure that all the restore points are recoverable at any point in time. I mean, since you have a health check scheduled to run periodically, then those previous restore points were also verified someday during previous job runs. Of course, there's a chance of storage bit rot and things like that so if you can justify the effort, you can perform additional checks of the entire chain.

Can validator run concurrently with an active backup job, or will it block the job like the health check while it's running?

Depends on the parameter it is run with. Parameter '/file' shouldn't lock the files.

[markus.wild]

Thanks!