PBS as a VM
- Thread starter cyruspy
- Start date
But PBS is just a fancy Debian. So you could try to use some other sync programs with Debian support and use them to sync the files of your datastore to a cloud storage provider.
That will be way slower (no incremental backups) and will probably need a multiple of the space, as Vzdump backups can't be differential. So backing up 10x 100GB will result in 1TB of backups and not just 150GB or so.
Last edited:
Ok, that may work. But in this case, I don't need 100% super ultra guarantee on the storage, because if it crashes I still have a copy in cloud. Then my original question remains: what is the best storage for for the PBS in a VM in this case? Just a regular virtual hard disk in a regular PVE LVM storage?
Yes, but won't require a full 4GB machine just for the backup. But indeed, I have to play with PBS a little bit to learn its potential.
But don't forget to enable the maintaince mode before syncing the datastore to the cloud and disable it afterwards again. So that the backup/restore/GC/prune/sync/verify tasks won't write to the datastore while syncing it.
Disk passthrough should be fine. ZFS would give a bit more data integrity (PBS can only detect bit rot, it can't repair corrupted data...ZFS in a mirror could also repair it) but then you would need another SSD, even more RAM and your existing SSD is most likely not a enterprise SSD, so not a good idea to use it with ZFS. So I would probably just use disk passthrough and then partition and format that SSD in the PBS VM with ext4 or xfs.
You don't need 4GB. Here it is working perfectly fine with a 2GB RAM VM. And I guess with PBS in a LXC or bare metal it can even be a bit less RAM. Like 1.5GB or so, as you don'T need to run the kernel.
Last edited:
I have rsync scheduled to run after my backups complete (cron job) up to an Ubuntu instance I have in Azure. It's syncing the backup datastore, but I'm wondering how that works for a restore? My PBS is a VM on PVE since I only have one server in my home lab. The backup datastore is on a different physical drive to the PVE and the ZFS pool with the other VMs that are being backed up.
On PBS is there a catalogue of what's been backed up that I should also backup to Azure and what's the process for doing a restore if my ZFS has a major failure?
You would need to do a new PVE installation, create a new PBS VM, get the filesystem with your datastore somehow into the new PBS VM and edit the datastore.cfg so you get a datastore with the sme name es before pointing to the folder of your old datastore. In case your local PBS disk survived you can then restore your guests. If not you would need to sync the whole datastore back before you could start any restores.
Ok, so there is no catalogue of what PBS has backed up, I just need to get the backups onto PBS and then I can restore them to my rebuilt PVE.
I’ve since gotten PBS working on an Azure Debian VM, so now I don’t need to use rsync and have a PBS sync job copying up to Azure. Only question I have left is how an I get an email notification from a sync job? I’m getting emails for my backup and verify jobs.