Function Verify backup

[simrsta]

I have installed PBS
and I'm happy because it's easier to backup and faster

My question is, do I have to verify all my backup results for every backup?

If it is not verified, what will be the impact?

because I do backups for the database server every 2 hours because it is related to patient medical record data

 

[Chris]

I have installed PBS
and I'm happy because it's easier to backup and faster

My question is, do I have to verify all my backup results for every backup?

If it is not verified, what will be the impact?

because I do backups for the database server every 2 hours because it is related to patient medical record data

Hi,
a verification job makes sure your backup snapshots are still intact. How often and when you want to check this is up to you and your needs. Further, this depends on other factors as e.g. storage setup as well, as the storage itself can have some data corruption detection and healing mechanisms (e.g. when using a ZFS RAID).

Re-verification of at least once per month is however recommended. Details you can also find in the docs https://pbs.proxmox.com/docs/maintenance.html#verification

 

[simrsta]

Hi,
a verification job makes sure your backup snapshots are still intact. How often and when you want to check this is up to you and your needs. Further, this depends on other factors as e.g. storage setup as well, as the storage itself can have some data corruption detection and healing mechanisms (e.g. when using a ZFS RAID).

Re-verification of at least once per month is however recommended. Details you can also find in the docs https://pbs.proxmox.com/docs/maintenance.html#verification

oh ok at least one per month

In PBS, backups are always incremental, right?
because I backup every 2 hours
so just add the latest data


i already backup in Synology but need 2 hour every backup and only 1x backup perday

 

[Chris]

oh ok at least one per month

In PBS, backups are always incremental, right?
because I backup every 2 hours
so just add the latest data


i already backup in Synology but need 2 hour every backup and only 1x backup perday

Yes, since the backups are performed incremental by deduplicating chunks, the verification makes sure the already uploaded and therefore reindexed chunks are still healthy. Otherwise all the snapshots referencing a corrupted chunk are corrupted.

If a corrupted chunk is detected, the same chunk is reuploaded if encountered on the local source, thereby healing all the snapshots referencing it. If the local data changed however the snapshots referencing the corrupt chunk are not restorable. So catching data corruption early is desired.

 

[simrsta]

Yes, since the backups are performed incremental by deduplicating chunks, the verification makes sure the already uploaded and therefore reindexed chunks are still healthy. Otherwise all the snapshots referencing a corrupted chunk are corrupted.

If a corrupted chunk is detected, the same chunk is reuploaded if encountered on the local source, thereby healing all the snapshots referencing it. If the local data changed however the snapshots referencing the corrupt chunk are not restorable. So catching data corruption early is desired.

How long does an incremental backup take??
It's been 10 minutes but it's still processing

I see that network traffic is not running but the backup is still in process?

 

[Chris]

View attachment 71063

How long does an incremental backup take??
It's been 10 minutes but it's still processing

I see that network traffic is not running but the backup is still in process?

Since this is a container, the PVE side will have to re-read and re-chunk the whole container filesystem, which depending on the data can take some time. Only chunks not known to the PBS host will be uploaded again.

There is an (still experimental at the time of writing) change-detection-mode which does allow to re-chunk only file payloads for files changed with respect to the previous backup run, further details for this can be found in this thread https://forum.proxmox.com/threads/pbs-client-change-detection-mode.150538/post-680876

 

[simrsta]

Since this is a container, the PVE side will have to re-read and re-chunk the whole container filesystem, which depending on the data can take some time. Only chunks not known to the PBS host will be uploaded again.

There is an (still experimental at the time of writing) change-detection-mode which does allow to re-chunk only file payloads for files changed with respect to the previous backup run, further details for this can be found in this thread https://forum.proxmox.com/threads/pbs-client-change-detection-mode.150538/post-680876

I make backups every 4 hours
but getting an error notification in prune
what do I have to do

 

[Chris]

I make backups every 4 hours
but getting an error notification in prune
what do I have to do

View attachment 71534

Well, the user you are running the backup job as does not have the permissions to prune backups on the PBS side. You will have to adjust the access controll for that user. See https://pbs.proxmox.com/docs/user-management.html#access-control

 

[luckman212]

I'm still a little confused about how to properly configure the verification jobs. I had posted before about yellow warning exclamation marks appearing next to everything in my repo before, but they came back. So I added a full verify job.

Does the way I have it configured here make sense?

I just added the weekly (sat @ 3:15am) job that does NOT skip verified snaps. The other job remains at "skip verify=YES and re-verify after 30d (which I assume will now never occur since the other job is weekly??) I tried to set it to "0" or blank but the UI does not accept that so I left it at 30.

 

[Johannes S]

Normally you would configure one job which will skip the verified job and reverify the skiped ones after some time of time and another less executed one for the full-verify.
As example I have one verify job which is run every day and will skip the already verified but still re-verify them after seven days:

And another one which will do a full verify at 4:00 am on the first day of the months:

Please note the difference in duration:


So in the end I have a daily verify of new backups and after seven days I can be quite sure that most of my backups are ok. The monthly full-verify is an additional safety that the backups are still fine but in theory shouldn't be needed.
Depending on your usecase and dataloss paranoia the schedules could be adapted e.g. the first one could be changed to do one weekly verify with re-verify after 30 days and the full-verify to every two or three months.

Regarding the question marks: They indicate, that a backups wasn't verified yet which is normal for new backups.
And it's normal that part of your backups are not in a verified state, the datastore has a setting whether new backups should be verified or not:



As you see I have disabled this, most likely yours is disabled too. Whether you enable it or not is kind of a tradeoff. If you enable it any new backup will be verified so you know directly whether it's ok or not. But if you don't enable it the backup job will be finished earlier. Another thing to consider: A regular verify job will only verify each chunk once, even if it's part of multiple backup snapshots.

As you propably know PBS works by splitting the backups data in a lot of small files (chunks) which are then referenced in each backup snapshot who should have them: https://pbs.proxmox.com/docs/technical-overview.html

The consequence is that each chunk will only be needed to upload and verified once, thus saving storage space and execution time for backups (an already uploaded chunk doesn't need to be uploaded again) and verification ( a backup job verifying multiple backups needs only to verify a chunk for the first backup it checks).