PBS verfication vs. zfs scrub

[Toxik]

Hi,

PBS documentation about backup verification says:

It is recommended that you reverify all backups at least monthly, evenif a previous verification was successful. This is because physical drivesare susceptible to damage over time, which can cause an old, working backupto become corrupted in a process known as bit rot/data degradation. It is good practice tohave a regularly recurring (hourly/daily) verification job, which checks newand expired backups, then another weekly/monthly job that will reverifyeverything. This way, there will be no surprises when it comes to restoringdata.

IMHO it would be a better idea to do a "zfs scrub" on the data store on a regular basis instead of re-verification of backups, wouldn't it? ZFS has everything included to save your data (and even fix it again, when on a redundant array).

What do you think?

 

[Chris]

Hi,

PBS documentation about backup verification says:


IMHO it would be a better idea to do a "zfs scrub" on the data store on a regular basis instead of re-verification of backups, wouldn't it? ZFS has everything included to save your data (and even fix it again, when on a redundant array).

What do you think?

Hi,
yes, but this is not mutually exclusive. However the datastore is not necessarily located on a zpool and should be storage agnostic. Therefore the wording of the documentation.

If you have the datastore as zfs dataset, then performing regular scrubs is definetly a good idea.

 

[fabian]

note that the verification will also catch problems other than bit-rot, such as admins accidentally deleting files, which ZFS cannot detect as an issue.

 

[Toxik]

Well, that makes sense