[SOLVED] Very long backup times

[Myze]

Hello,
I have a promox server with two proxmox backup servers.
Since the second one is in place, backups take at least 4 or 5 hours instead of a few minutes with only one.
Have you ever experienced this?
Thank you for your help,

 

[UdoB]

Every time you switch the destination the source has to be read again, completely.

That's why backup to one PBS and then sync to the other is much more lightweight for the source than backup to two PBS in an alternating manner.

 

[Myze]

Ok thanks for the info, isn't there any way to make him understand that he doesn't have to take the whole thing back?

If not, I'll configure the backup to pbs1 and then the synchronization from pbs 1 to pbs 2.

Thanks

 

[UdoB]

Ok thanks for the info, isn't there any way to make him understand that he doesn't have to take the whole thing back?
If not, I'll configure the backup to pbs1 and then the synchronization from pbs 1 to pbs 2.

As long as you sync to the same PBS-datastore: only chunks that are not already there will get transferred from the source. In my understanding it does not matter what the source is.

Watch "Storage usage (bytes)" when you run the first sync. It should not increase much.

PBS is clever

 

[Myze]

I set up the synchronization job on pbs1 after emptying the datastore on pbs2, but the synchronization doesn't transfer anything. I must have missed something?

 

[UdoB]

I set up the synchronization job on pbs1 after emptying the datastore on pbs2, but the synchronization doesn't transfer anything. I must have missed something?

Wrong direction? Sync does pull data. A job on pbs1 to sync with pbs2 would pull the empty datastore from pbs2 to pbs1.

https://pbs.proxmox.com/docs/managing-remotes.html#sync-jobs

 

[Myze]

Yes, that's right, thank you, I hadn't seen that.
The synchronization seems to be running now.

 

[AmateurProgrammer]

In my opinion it could be nice to have an option in the pbs to use ceph fs for example and have the same datastore available from multiple pbs instances. The different PBS servers could have some sort of locking mechanism on the datasets. That would enable kind of clustering for the PBS system.