PBS as a target for PVE and additionally OFF-Side

[pixel24]

Hi@all,

I have a PVE (current version) on which several VMs are running. I do not use ZFS but hardware RAID + LVM. PVE saves the dumps to an external USB hard drive.

However, I would also like to have these dumps OFF-side. In case the roof burns.

Can PBS be a solution for this?

with best
pixel24

 

[pixel24]

Can PBS also host virtual machines like PVE? I need the possibility to run VM's on the second server (in the LAN) because I need a UCS with Bareos for our "normal" backup.

Or can I set up PBS as a VM on PVE?

At home I have a Synology NAS that is accessible from outside. Is it possible for PBS to write selected discs that it has previously backed up from PVE to this external destination?

 

[UdoB]

Or can I set up PBS as a VM on PVE?

Yes, sure. Why not?

And you also/alternatively can install PBS Server on an additional PVE Node - in the base OS. I did it this way because it increases the number of cluster-nodes = it helps keeping good quorum and it can work as an additional compute node if necessary (and if that node has enough compute power and ram).

Best regards

 

[VictorSTS]

Your best bet for offsite backups is to have one PBS server and another PBS server at the remote location. Then, configure replication within PBS, so the repos you want to be offsite are periodically replicated to the offsite PBS. Any backup you add to the replicated repo will be eventually synced to the remote one too, so its easy to maintain.

Other, not recomended, option would be having TWO local PBS instances, your current one and another PBS which will somehow connect a remote filesystem (i.e. NFS, iSCSI) and set up replication among both PBS. Performance will be worse and ready yourself to some troubles if the remote filesystem is not available for any reason.

 

[pixel24]

Thank you very much for the answers!

And you also/alternatively can install PBS Server on an additional PVE Node - in the base OS.

Do I understand correctly that I can install PBS as "software" directly on PVE? That is, without an extra VM?

Your best bet for offsite backups is to have one PBS server and another PBS server at the remote location.

ok, that sounds reasonable and I will do it that way. The external PBS does not need as much power as the internal one. I think.

 

[UdoB]

Do I understand correctly that I can install PBS as "software" directly on PVE? That is, without an extra VM?

Yes.
After adding the repository of the PBS you can install it straight forward. The result on my cluster-integrated but for-backup-only node:

root@pveb:~# pveversion && proxmox-backup-manager version
pve-manager/6.3-6/2184247e (running kernel: 5.4.103-1-pve)
proxmox-backup-server 1.0.8-1 running version: 1.0.8

I am not sure if this is really a good idea for every situation. But I run Proxmox in a pure homelab - and for me this combination is great!

Best regards

 

[pixel24]

The idea would be this:

Main server with PVE to run the VM's. This already exists.

I install PVE with PBS on the backup server.

If the main server fails, I can make the VM's available on the backup server relatively quickly.

Externally, I would take a "normal" (somewhat stronger) PC and install PBS.

 

[VictorSTS]

I am not sure if this is really a good idea for every situation. But I run Proxmox in a pure homelab - and for me this combination is great!

Best regards

It is never a good a idea for any kind of production environment, even at home, to have both your data and backups in the same box. It that box burns you will lose it all. At least have the PBS repo in a usb drive that you can connect to another PBS to recover your data.

 

[UdoB]

It is never a good a idea for any kind of production environment, even at home, to have both your data and backups in the same box.

Yes, that's true for most scenarios. And that's also why I put emphases on "for-backup-only node", "not sure" and "for me"

Best regards

 

[pixel24]

Hi@all,

I would like to address this issue now. But I still have a question about the OFF-side.

The environment:

Here in the LAN a main server (PVE) on which the VMs run. Also in the LAN, a backup server PVE+PBS. This serves as a backup target for the main server.

The backup server is therefore with PVE+PBS so that I can run the VMs there in the event of problems on the main server.

I hope I have understood everything correctly.

Now to the OFF-side. There (via WAN) I would place a server with PBS which serves as a mirror for the backup server.

The current VZDUMP is 1.5TB and is backed up every night. How does PBS work? Does it transfer the whole amount every night or does it transfer differences of the dump? 1.5TB is a bit much for a 30MBit (LAN upload) line.

with best
pixel2404

 

[leesteken]

A similar question about using external drives to move the backup off-site was discussed (and actually tried by a user) on the pve-user mailing list.

 

[pixel24]

I read the post but there it goes to an external hard drive. I want to "mirror" the PBS (LAN) over the Internet to a PBS (EXTERNAL).

But what I learned from the article is that it is better to run the PBS as LXC on a PVE.

 

[leesteken]

I read the post but there it goes to an external hard drive. I want to "mirror" the PBS (LAN) over the Internet to a PBS (EXTERNAL).

My apologies, I interpreted your remark about upload speed looking for another method of data transfer. Note that a PBS sync only has to transfer new chunks (which roughly translates to the data that changed between syncs), which can help a lot if there is not too much change or many changes are identical.

 

[pixel24]

That sounds good. So PBS does not transfer the whole VM to the external PBS every time, but only the changes it contains that have happened since the last sync?

 

[leesteken]

That sounds good. So PBS does not transfer the whole VM to the external PBS every time, but only the changes it contains that have happened since the last sync?

I'm sure other people can explain it better, but PBS always creates a full backup of a VM by reading and compressing all the data and storing it in chunks. Every (full) backup is just a list of those chunks. If there are no (or little) changes between backups then all (or most) chunks are the same and do not need to be stored again (just the list itself is slightly different). This is called deduplication and even works over different (but very similar) VMs (when using the same encryption key). This also means that to sync a PBS with another, only the lists (that describe the contents of full backups of VM) and any missing chunks need to be transferred.

 

[pixel24]

Ok, thanks for the info. I'll just set up a test environment now to get a first impression of how it works