Dataset and qcow2

Hello everyone!

I have a question about creating datasets and qcow2 disks on Proxmox. Instead of using zvols, which Proxmox uses by default as ZFS storage, I create a dataset that will act as a container for all the disks of a specific virtual machine. Then, in Proxmox’s storage settings, I map a directory (which is a type of storage in Proxmox) and create all the virtual disks needed for the virtual machine.

I have often heard from @mercenary_sysadmin that creating datasets and raw or qcow2 disks is difficult because it is not natively supported. My procedure is certainly not as straightforward as the point-and-click creation of zvol storage by default, but it is not complicated.

At this point, I would like to know if @mercenary_sysadmin is referring to this procedure or something completely different?

1 Like

If you found a way that’s not an absolute ball-ache, I’m more than happy for you, and honestly I kinda wish you’d write up a tutorial on how to do it. I’m not a Proxmox user myself, I’ve just been told by countless Proxmox users that it’s incredibly difficult to get it to use file-based back ends for VMs without resorting to popping down to the console to do it. :slight_smile:

You can absolutely do that. What you are looking for is a Proxmox Directory.

Basically you tell Proxmox where to store qcow2 files. It’s easy but you need to go to the CLI.

1 Like

The obnoxious thing is that you have to repeat that with every single VM you create, because otherwise you wind up with all your VMs in a single dataset (where they can’t be rolled back independently from each other).

Totally right on that one. Super annoying lol

I confirm that it is necessary to create a dataset for each virtual machine (and additionally a storage directory) in order to properly manage snapshots on each individual VM. However, I don’t see it as very different from what I do on other servers with Debian and KVM. Ultimately, for each new VM, you always need to create a dataset with the terminal and map it to a directory (if using Proxmox) or to a pool (if using KVM/libvirt). Therefore, I mentioned that I don’t see anything complicated or out of the ordinary regarding managing datasets and qcow2 files on Proxmox.

Am I missing something?

I think the thing that you’re “missing” is that if you need to constantly jump down to the CLI, you’ve invalidated most of the point of using Proxmox in the first place. If you’re willing to do largely CLI-based management, you would typically be better served with something like vanilla Debian or Ubuntu that’s more widely supported, offers more flexible options, and can do far more things.

Essentially, the only selling point Proxmox has is its GUI, so if that GUI doesn’t work for you, there’s not much reason left to use Proxmox instead of a vanilla distro.