Have I set something up incorrectly or was I under a completely different impression about how PBS works?

[ikidd]

I set up PBS a while ago for testing, and had it backing up my 2 PVE servers. These PVE servers are using ZFS of course, and so I set up PBS to use a ZFS pool as well.

I ran up against storage issues, so I started looking to see which datasets aren't compressed and/or dedup hasn't been enabled. I look into the shell on the PBS server, and I don't see any datasets and snapshots. I would have thought the integration of ZFS would imply that the PVE servers were doing a ZFS send of snapshots to the PBS server, with all the features that that would imply, and that remotes sycing to that PBS server would do the same. Instead, it looks like it's doing a standard block copy into one datastore on the PBS server.

Did I set something up incorrectly, or does PBS just not actually use ZFS datasets like a backup solution using Syncoid would? Because that seems extremely backwards for an integrated backup solution for a stack that seems so targetted at ZFS storage. I have to figure I just messed up something in the config and can't read a manual.

 

[Dunuin]

First is PVE not ZFS targeted. It allows alot of storages and something like LVM-Thin or CEPH might even be better in other scenarios. And from what I read here in the forums the majority isn't using ZFS. For homeservers people mostly choose LVM-Thin because it can run on cheap low-end consumer hardware without big overhead. And bigger companies that need a HA cluster of PVE nodes will use CEPH as it got all the features of ZFS but is better scalable and works as a real shared storage. ZFS is more for the cases in between, so more used for SOHO when you just want 1 or 2 reliable nodes that use enterprise grade hardware.
And PBS doesn't make use of any ZFS features. Something like a ext4 would be totally fine for a datastore as PBS does zstd-compression, deduplication, checksumming and verification, aes encryption, syncing and so on all in software so no such filesystem features are actually needed.
That way it can be used to backup all guests no matter on what storage they are stored on.

 

[ikidd]

I see.

Thanks.