Backup Server as a VM and pass through drive

[hermannw]

Hi,

I am quiete new to Proxmox, so my question might be easy to answer.

My current home environment consists of a single Hyper-V host with a few Windows VMs, one of which is critical.
The entire system is optimized for energy efficiency and runs on a Mini-ITX board. Performance is not the goal.

For backup, I use Vembu, which is installed on Hyper-V. That's not best practice, but it has been working without any issues for over five years.
The backup is scheduled weekly using an external hard drive via eSATA as backup target. Switch on the HDD will mount it to the Hyper-V and makes it available to Vembu.

I would like to migrate the VM to Proxmox. I have already tested this, and the migration process works.
For backups, I would like to use Proxmox Backup Server. Although not recommended, I intend to install it as a VM.
I assume it is also possible to pass through drives to a VM (RDM) via eSATA or USB.

The goal is to have scheduled backups configured so that, at the respective times only the external hard drive needs to be switched on.
I would like to avoid having another physical machine running.

Will this work reliably, or does anyone have experience with such a setup?
Or is there a different approach?

Thanks,
Hermann

 

[UdoB]

Will this work reliably,

It may work for you. Every single backup is better than none.

Personally I would not accept this approach as "okay, Backup done" - especially if it is your only backup.

Obviously you lose the backup mechanism when the only host dies. Sure, the data is on that external drive. But:

• that external drive is a single drive, right? So... no redundancy --> bad design (from my personal point of view!)
• it is a SSD, right? If you only have a single spindle of rotating rust I would recommend to use "vzdump"-based backups instead of a PBS as PBS needs IOPS!
• after a disaster you need to install PVE first, from scratch. Then re-install PBS from scratch, as you don't have a backup you can access now

Of course you would probably do this while under stress as your whole virtual world is defunct. This aspect might be a larger problem than one might expect. Been there; done that.

And then there is the 3-2-1 rule: have at least three backups, on two physically different media, one of it stored off-site. People have had bad experience when ignoring this approach.

Whatever you do: remember to test "restore" after a (simulated) disaster --> until you actually did this you only have a "Heisenberg-Backup". (( Edit, better: "Schroedingers-Backup" = it is alive/usable and/or just dead/trash when you're going to restore it... ))

----
Personally I have a total oversized cluster in my basement, so it is clearly not comparable to you. Just as a glimpse on what I've done for myself: a single PBS is running 24*7. It is small but fast. It is integrated in my cluster = no separate hardware but living on a cluster node. Then I have multiple older computers with some large volume as rotating rust (always plus a fast "Special Device"). Some of these secondary PBS' are turned on automatically (only) once per week. Some are turned on manually once per month, in a round-robin fashion. This is for both energy saving and to "air-gap" them most of the time.

Have fun!

 

[hermannw]

Udo,

Thank you very much for your detailed explanation.
My approach differs from yours as I prioritize low power consumption over.
That is why my entire setup (server, router, switch, UPS) consumes a mere 32 watts.

I can fully endorse your points regarding backups as I followed the same practice in professional environments.
However, this is my home setup, which I back up once a week, including restore tests of course, as well as a full restoration of the Domain Controller on one occasion. This is clearly not a "okay, Backup done" approach.

So, it seems, that I should take a closer look at `vzdump`.

The reason I asked about disk passthrough was that, back in the days I used XenServer, this frequently caused various PBD issues.
Passthrough, provided it works reliably, would still be my first choice.

Best,
Hermann

 

[_gabriel]

Here all PBS install are alongside PVE.
Disk dedicated to PBS datastore.
Single HDD for few VM is enough
vzdump is only good for manual/one shot backup.
For scheduled backups PBS is mandatory.
Speedy and Verify schedule.
Restore is faster since PVE versions later 2025.
Not recommended because can be error prone, but once you learned, tried few times, how to mount/attach/connect PBS datastore, you can't leave it.