As someone who’s just bumping up against the limitations of lxd/incus, this is very interesting, thank you! Do I take it it’s using the default nixos kernel?
Really exciting to see this, as this has been one of the largest pain points remaining in my network - given its the only device without declarative configuration!
If you need beta testers and have a Matrix room or similar I’d be happy to try things out if that’s helpful.
One question I have is the mentions in the README that you need to set configuration declaratively and also in the WEBUI “to keep Proxmox happy” - could you expand on that?
In other words I could just rely on the nix config to stand up an exact replica server, it’s just that the WEBUI wouldn’t show the applied config?
Currently (and we might fix ça in the future) the Proxmox web UI cannot detect the interfaces that exists on your system from your NixOS configuration. That poses a problem when you try to add a virtual network interface to a VM through the UI because then it doesn’t list the bridges that you could connect it to. The fix here is simply to create a bridge through the web interface (you only have to do it once) and then Proxmox will think it is handling the configuration of this bridge (while it is actually your NixOS configuration).
Are you guys planning to upstream this into nixpkgs at some point? I understand it’s experimental right now. But I personally run Proxmox quite a bit, and I’d love to see this become really usable in production. Which for me means that I would not have to rely on a third-party cache.
Exciting. What “balance” of iterative/declarative configuration are you targeting, if any? For example, will it be possible to declare a full system as per usual, but allow iterative changes via WUI/CLI that are restored on a reboot?
Im definitely considering using this.
Is this home-lab ready?
I use proxmox and have quite liberal use of the LVM / PBS, etc.
I would use microvm if I had a web interface so im kinda stuck on proxmox.