Proxmox Backup server deployment options on PVE

[plokko]

My network is composed of two PVE servers:
one primary server where all the VM runs and one secondary backup server used to store VM backups and as a backup PVE (in case the primary server fails it must be capable of restoring and running the VMs).
Both servers runs on ZFS, all the disk on the machines are used for PVE (and VMs).

I need to make the backup storage of the secondary server accessible both from the local PVE and from the primary server;
as PVE does not allow to share disk from the GUI i have to use a service like PBS (proxmox backup server).

There are three way to run PBE:
1. Installed on the PVE:
very easy to do, shares all the PVE resources with minimal drawbacks, no need to change anything as they share the same logins and disks but can be accessed on different ports on the same IP.
May not be ideal as it's integrated on the hypervisor.

2. As a VM on the secondary PVE server:
Allow more separation as the service is encapsulated on a VM, the drawback is that the local PVE does not have direct access to the data and must relay on the VM ( uses resources to access data, if the VM is off or down i cannot access the backups, data resides on a virtual disk).
Note: i cannot dedicate the disk to PBE as they are shared with the local PVE.

3. Installed on a Debian CT:
Allow more separation, the PVE backup folder may be shared with the CT (bin mount) so that if the CT to allow direct access to the data from the PVE even if the CT is off.
The disk and ZFS status are not readable even in privileged CT (SMART status,etc.) but PVE does have the same interface so it's a non issue.

What's the best way to run PBE in this scenario?

 

[dylanw]

Why not combine options 1 and 2, and install it alongside your secondary PVE server?
The problem with option 1 is that if your system fails, your backups may become inaccessible.
With option 2, as you mentioned, you have all the content is in a virtual disk, and any issues with the VM could be problematic.

 

[plokko]

Why not combine options 1 and 2, and install it alongside your secondary PVE server?
The problem with option 1 is that if your system fails, your backups may become inaccessible.
With option 2, as you mentioned, you have all the content is in a virtual disk, and any issues with the VM could be problematic.

In case of hardware fail (machine with disk intacts or unable to run local PVE) with option 1 i can put the disks on another system (only a system capable of reading ZFS is needed) and extract the backups even if the PVE is incapable of running;
with option 2 i have to extract the disk image from the ZFS (it's also not even saved as a file) and run it on a PVE machine, then extract the backups from it.

In case PBE became inacessible with option 1 i still have access to all my backups from the PVE, i just have to copy them or make them accessible again (fix PBE, create a CIFS share or move to a shared folder for example);
with option 2 i have to fix the PBE environment or mount the virtual disk and extract the backups from it.

So even if i'm against putting anything on the hypervisor the drawbacks of putting it on a VM are more than the benefits.

What did you mean by "combining option 1 and 2"?

 

[dylanw]

What did you mean by "combining option 1 and 2"?

Sorry, that was probably a weird way to phrase it... I just wanted to say that you could install proxmox backup server on the same system as your secondary PVE server, rather that as a VM on the secondary server. Doing this would give you the simplicity you specified in option 1, without some drawbacks from option 2 (such as having all the data stored on a virtual disk, locked inside a VM).

I thought this would be best, as it means your backup system is totally separate from the data being backed up, the raw data is still accessible from the (second) host system (also helpful in case you need the physical disks), and VMs can be immediately restored on to the second server without having to move disks around in case of failure in the first.

with option 2 i have to extract the disk image from the ZFS (it's also not even saved as a file) and run it on a PVE machine, then extract the backups from it.

In any case, couldn't you just use qmrestore and pct restore, to restore the VM/CT on to the first/second machine? [1]

[1] https://pve.proxmox.com/pve-docs/pve-admin-guide.html#vzdump_restore

 

[plokko]

Sorry, that was probably a weird way to phrase it... I just wanted to say that you could install proxmox backup server on the same system as your secondary PVE server, rather that as a VM on the secondary server. Doing this would give you the simplicity you specified in option 1, without some drawbacks from option 2 (such as having all the data stored on a virtual disk, locked inside a VM).

I thought this would be best, as it means your backup system is totally separate from the data being backed up, the raw data is still accessible from the (second) host system (also helpful in case you need the physical disks), and VMs can be immediately restored on to the second server without having to move disks around in case of failure in the first.


In any case, couldn't you just use qmrestore and pct restore, to restore the VM/CT on to the first/second machine? [1]

[1] https://pve.proxmox.com/pve-docs/pve-admin-guide.html#vzdump_restore

Ok, thanks for the clarification, that was what i had in mind in option 1.

For the restore procedure yes, it's not impossible or hard, but the first option may be just a little bit easier