VM or PVE for pbs
- Thread starter Arthur777
- Start date
Recommended would be none of those and to get a dedicated server just for PBS with local SSD storage. Or even better two with one offsite and then kept in sync via sync tasks.
PBS next to PVE bare metal got the benefit that both could share the physical hardware as well as resources and that there is no virtualization overhead (like no nested filesystems required). Downside would be the added dependency compared to a VM. So if for example PVE9 gets released you can't upgrade your PVE8 to 9 until PBS also gets an major upgrade from PBS3 to 4. Both variants won't help you in case of ransomware or human error. For that you really want a dedicated server just for PBS that isn'T doing anything else (or at least nothung related to your PVE nodes/clusters.
In general not a good idea to store your backups on the same server that running the stuff you want to backup. Lose that server for whatever reason and not only your VMs/LXC but also the backup of them will be lost.
And running PBS on the PVE host but backup storage remotely on some NAS isn't either. PBS needs high IOPS performance (thats why you should backup to local SSDs and should avoid using HDDs) so you don't want that additional network latency/overhead and in contrast to VZDump backups you can't do anything with those backups on the NAS without a working and running PBS.
PBS next to PVE bare metal got the benefit that both could share the physical hardware as well as resources and that there is no virtualization overhead (like no nested filesystems required). Downside would be the added dependency compared to a VM. So if for example PVE9 gets released you can't upgrade your PVE8 to 9 until PBS also gets an major upgrade from PBS3 to 4. Both variants won't help you in case of ransomware or human error. For that you really want a dedicated server just for PBS that isn'T doing anything else (or at least nothung related to your PVE nodes/clusters.
In general not a good idea to store your backups on the same server that running the stuff you want to backup. Lose that server for whatever reason and not only your VMs/LXC but also the backup of them will be lost.
And running PBS on the PVE host but backup storage remotely on some NAS isn't either. PBS needs high IOPS performance (thats why you should backup to local SSDs and should avoid using HDDs) so you don't want that additional network latency/overhead and in contrast to VZDump backups you can't do anything with those backups on the NAS without a working and running PBS.
Last edited:
Hello, when you say this, would it be correct to have three PVE in a cluster: one dedicated to PBS, and two others for installing VMs? Plus, an additional PVE for an offsite PBS
I dont really understand your point to be honnest
Thanks for all of your advices
If for example a new major PVE version will be released 3 months before the next major PBS version, you won't be able to upgrade your PVE for 3 months. So not great if you want to be up-to-date.
You also can not first upgrade PVE, check if the PVE upgrade won't break anything and only then, after verifying this, upgrade your PBS. You will have to upgrade both at the same time. If there is then a problem (hardware not compatible anymore or whatever) you won't be able to run your VMs and you also won't be able to access your backups anymore to be able to wipe PVE for installing the old working version again.
2x PVE (with no PBS) + 1x PBS as qdevice (with no PVE) in a cluster + 1 PBS offsite (with or without PVE but not clustered) would be fine.
That would be the minimal setup for HA and ransonware protected PBS with offsite backups.
