[PVE-User] Revisited: External disk backup using PBS - Requesting Criticism/Advice?

[Arjen]

On Saturday, April 10th, 2021 at 05:22, Lindsay Mathieson <lindsay.mathieson at gmail.com> wrote:

> TTL;DR - Backup PC is a standalone Proxmox Server, running a PBS lxc
> container using its host filesystem for the Backup store. The PBS
> container is backed up using vzdump to an attached external hard drive
> which is rotated offsite.
>
> Source Data
>
> -   5 Node Proxmox Cluster
> -   Ceph storage (size = 3)
> -   30 VM's
>
>     Backup Destination
>
>     Standalone Proxmox Server in same server room as main cluster
> -   16 GM Ram
> -   CPU: Intel i5
> -   Boot - 2 SSD's in ZFS RAID1
> -   Data - 2 4TB WD NAS Drives in ZFS RAID1
> -   Bonded 1G * 3
> -   PBS Container
>
>     o Data Store - 4TB, passed through from Host
>
>     Schedule
> -   Proxmox backups up all VM's to PBS Container Weekly
>
>     o Will revisit the schedule
> -   Host proxmox server backs up PBS and its data to external hard drive
>
>     o Approx 2 TB Data (we are just an SMB)
>
>     o Drive is rotated offsite
>
>     Recovery
> -   VM's can easily be restored from the PBS server as needed (very rare
>
>     occurrence - usually a user messed up their VM)
>
>     Disaster Recovery
>
>     This is the real concern - Fire, Theft etc. All servers and data
>     including the Standalone Host and PBS server are gone.
> -   Recreate Proxmox Cluster
> -   Recreate Proxmox PBS Host
>
>     o Restore the PBS Container and data from an offsite backup disk
> -   Restore the VM's to the Cluster from the PBS Container
>
>     Does all this seem practical and safe?

Don't expect to be able to backup the PBS container with 4TB to a 2TB external drive. The Datastore of a PSB does not compress much further and Proxmox VE Backup will only backup virtual disks and not mountpoints or storages passed from host, if I understand correctly.
I suggest adding a virtual disk of 2TB to the PBS container (and format it with ext4) which can be backed up by the Proxmox VE Backup.
I would also run the PBS container (with virtual disk) on the cluster instead on separate hardware which is a single point of failure. The local PBS would be then just as reliable as your cluster.

Regarding safeness: I suggest doing a automated disaster recovery every week to make sure it works as expected. Or at least partially, like restoring the PBS from an external drive.
Regarding practicality: I would have a remote PBS sync with your local PBS instead of moving physical disks (but you mentioned before that that was not really possible).

best regards, Arjen