After seeing Backup Server go GA, I was immediately interested in putting it to use in my home/lab environment.
Currently my storage server backs up to warm storage over sftp with Duplicacy and it has some amazing compression+deduplication. It's also cross-machine deduplicating so any machines backing up to the same backup storage share the same deduplication table for optimal space savings.
With PBS officially out of beta, I wanted to move over to something like the below. Unfortunately the warm storage host cannot be made to be the backup server because it runs on ARM64 architecture.
My thought would be to make my current storage server a PVE host, run PBS as a VM and split my file shares out into their own VMs backing up to PBS. As for the warm storage I recognize that it could be painfully slow, but I was thinking of adding it to the PVE host as an NFS (or potentially CIFS) datastore. That should help alleviate know caveats of PBS and network datastores. I am already in the process of doing the above minus PBS.
The questions that come to mind however are:
1.) Is deduplication shared across all backups to the same datastore, or is deduplication per VM?
2.) How does PBS handle a disk becoming unavailable in the even that the NFS host lost network connectivity with the PVE host?
3.) How does PVE handle NFS connection problems (asking because I haven't had any in my lab yet
)
4.) Are there any caveats or issues with the above that I'm not thinking of?
I am planning on running routine verification jobs, so my hope would be that any data corruption caused by the PBS datastore being on a remove virtual disk will be caught at that point. The warm storage is on-site, so not cloud like the icon may make it seem. But it's a limited functionality host.
Thanks for your help Proxmox Gurus
Currently my storage server backs up to warm storage over sftp with Duplicacy and it has some amazing compression+deduplication. It's also cross-machine deduplicating so any machines backing up to the same backup storage share the same deduplication table for optimal space savings.
With PBS officially out of beta, I wanted to move over to something like the below. Unfortunately the warm storage host cannot be made to be the backup server because it runs on ARM64 architecture.
My thought would be to make my current storage server a PVE host, run PBS as a VM and split my file shares out into their own VMs backing up to PBS. As for the warm storage I recognize that it could be painfully slow, but I was thinking of adding it to the PVE host as an NFS (or potentially CIFS) datastore. That should help alleviate know caveats of PBS and network datastores. I am already in the process of doing the above minus PBS.
The questions that come to mind however are:
1.) Is deduplication shared across all backups to the same datastore, or is deduplication per VM?
2.) How does PBS handle a disk becoming unavailable in the even that the NFS host lost network connectivity with the PVE host?
3.) How does PVE handle NFS connection problems (asking because I haven't had any in my lab yet
)4.) Are there any caveats or issues with the above that I'm not thinking of?
I am planning on running routine verification jobs, so my hope would be that any data corruption caused by the PBS datastore being on a remove virtual disk will be caught at that point. The warm storage is on-site, so not cloud like the icon may make it seem. But it's a limited functionality host.
Thanks for your help Proxmox Gurus