Home Assistant LXC = extra work?

Running Proxmox on my server, with HAOS as a VM and Jellyfin as a LXC.
I see my HAOS VM is eating up all the 8 GB of RAM i’ve allocated to it, probably because of Frigate. I just spend the last 4 days trying to passthrough my iGPU to Jellyfin (as a VM) and only succeeding because i deleted the VM and started over with a LXC and Jellyfin running in Docker. I also wish to passthrough the iGPU to HAOS, but

  • I don’t know if it’s even possible to passthrough an iGPU to multiple VM’s and/or containers
  • Even if it is possible to passthrough the iGPU, i just spend 4 days and a plethora of guides, all with different approaches, trying to make sense of it all. With this experience, i’m pretty sure no one guide can make this work
  • since a lot of the passthrough work is done on the host, i’m a bit afraid of breaking the already working passthrough to the Jellyfin LXC

So i though about spinning up a HA LXC from TTECK but saw that you cannot do it with HAOS… I’ve never run HA as anything other than HAOS and i love having it all (Add-ons, backup, restore, managed OS) collected in one machine. I can’t imagine how much of a hassle it would be to run HA Core instead, with self-management of add-ons, making everything communicate properly, updating, backups and probably other problems/limitations that i haven’t thought of yet. And yes, i’ve seen the link to the installation comparisons, but that doesn’t answer my question.

If anyone has been in a situation like this or researched it, i hope you will share your experiences

I run HASS in a VM via pip and its a headache to update/fix.

Don’t get me wrong the others on here are a great help, but still if I didn’t need a few edge cases (private DNS, ssh it to other machines as automactions and a few other things I can’t remember right now) I wouldn’t be running it like that.

That said have a read of the pip install method and give it a spin, you’ve got nothing to lose testing it out.

I’m “passing through” my GPU to jellyfin and a few others. As LXC is less like a virtual machine and more like a container for services, the host is in charge of the GPU and allows programs in the LXC to use it, so can share resources.

It is not more like that. It is exactly that. :wink:

1 Like