Backup to NFS or PBS

[BurgelOutfox]

Hey everyone. I have a 3 node cluster running just fine. I have an older machine I could make another node to just play on. Or right now I set it up as a PBS just to see. But my cluster already backs up to nfs. And nfs backs up to offsite.

So I mounted the NFS and made a local and nfs datastore and setup sync job and such. So the cluster did a backup to PBS and PBS synced to NFS. So all works well.

Is there any benefit besides adding complexity for fun to put pbs in the middle?

Or make it a “testing node”

Thoughts or advice?

 

[bl1mp]

Hi BurgelOutfox,
the PBS provides deduplication. So unless your storage underneath the NFS is capable to prrovide the same, you nicht save some space.
That of course also depends on retention times and data structure. The system should probably also fit with your Recovery procedure.

BR, Lucas

 

[Johannes S]

So I mounted the NFS and made a local and nfs datastore and setup sync job and such. So the cluster did a backup to PBS and PBS synced to NFS. So all works well.

There are two problems with that approach:

• First network shares are not great with PBS to begin with see https://forum.proxmox.com/threads/datastore-performance-tester-for-pbs.148694/ Where do you host your nfs? If it's vm capable NAS it would propably work better to have the PBS as VM directly on the NFS. Just one example: For garbage collection and verify the PBS need to access (and in part read) all files of the datastore over the network. If it's on the same host garbage collection and verify tasks will be a lot faster
• PBS has a native sync mechanism and (beginning with PBS 4.0) also the ability to sync to s3 cloud storage. The s3 feature is still "technology preview" so I wouldn't use it as sole backup mechanism but I expect it to stabilise during the next year. The native sync to another PBS is rock solid and works quite well. On the other hand there were several reports in this forum where people messed up their backups by using something else (e.g. rclone) for offsite backups: https://forum.proxmox.com/threads/datastore-synced-with-rclone-broken.154709/ There are ways to avoid this problem but I wouldn't recommend that if you don't know what you are doing

So back to your question if you can't host as PBS as VM on your nfs server I would prefer to use your older machine for it. If you prefer to save on energy costs you could setup a PBS vm in your cluster to have fast restores, wake up your PBS every night, syncing the backups to it and power it down again. If you would like to have the PBS as additional virtualization host you shouldn't add it to the cluster but can install PBS and PVE to it. I wouldn't use this for anything important (backups and production on the same machine are a big no), but for trying things out that might be ok. And you also have the benefit, that in case your cluster gets broken you could restore your most important vms/lxcs to the PBS as "emergency-host"