[SOLVED] Can I have two PBS instances point to the same NFS Datastore location?

[PythonTrader]

I've installed and configured an instance of PBS using a NFS Datastore. I can back up my VMs with it and it is working well.
I am happy with this setup, since my NFS is based on a reliable instance of Synology.


Before I leave this setup to autopilot and move on, I like to simulate a disaster recovery situation so that I know when something really bad happened, I can create a new instance of PBS and restore my VMS.

For disaster simulation, can I install a second instance of PBS, point its datastore to the same NFS location and test the restore operation to a separate instance of PVE?

This would give me confident that I can rebuild my VMs anytime.

 

[Dunuin]

For disaster simulation, can I install a second instance of PBS, point its datastore to the same NFS location and test the restore operation to a separate instance of PVE?

It's no problem to use a datastore with one PBS and then replace it with another PBS with. But I can't image that it would be a good idea to run both PBSs at the same time, so make sure to disable stuff like autostart and so on in case PBS is a VM. And there is always the risk that the new PBS could prune/GC your datastore if you for example forget to set proper pruning tasks that fit the old PBSs configs.

 

[PythonTrader]

It's no problem to use a datastore with one PBS and then replace it with another PBS with. But I can't image that it would be a good idea to run both PBSs at the same time, so make sure to disable stuff like autostart and so on in case PBS is a VM. And there is always the risk that the new PBS could prune/GC your datastore if you for example forget to set proper pruning tasks that fit the old PBSs configs.

Hi @Dunuin , thank you so much for the informative answer, as always.


At this point my PBS+Synology NFS is doing the backups very well. My Synology has a 3-2-1 backup so I am confident that my backup files ARE SAFE.

Do you think is it 100% safe to move on and don't spend more time on testing and simulating disaster recovery?

My plan for disaster recovery is to build a new PVE cluster, add a new PBS, add the the same Synology NFS location to it and restore the whole cluster using the new PBS.

My cluster does not have large VMs but it has many small VM or CT that I spent so much time to configure and setup.

 

[fabian]

you should definitely do a restore test! having backups but not testing them is often not much better than not having backups at all - after all, if you don't check that they contain everything you need, you won't find out until it is too late..

although you can probably skip the "install a new PBS" part, the datastore itself is self-contained, as long as you know how to point a new PBS instance at it that part shouldn't be a problem.

edit: if you want to also test restoring PBS itself from scratch, you could just power down your original PBS for the duration of the test..

 

[Dunuin]

Do you think is it 100% safe to move on and don't spend more time on testing and simulating disaster recovery?

I already dropped an old PBS VM with datastore on NFS and replaced it with a new PBS LXC reusing the old datastore. So I can confirm that this can work.
But of cause always a good idea to test desaster recovery first. Then you are trained and know what to do when you actually need it.
You can also set the datastores maintainence mode to "read-only" if you just want to test a restore with a new PBS without risking to screw up stuff when later switching back to your old PBS.

Also don't forget to create a backup of your PBS. For example with Clonezilla to your NAS if it is bare metal or using VZDump to your NAS in case it is virtualized. So you can quickly restore the whole PBS.
Also makes sense to backup the /etc folder (especially /etc/proxmox-backup) of your PBS so you got a copy of all the PBS configs.