MAAS, LXD, K8s and GPU passthrough

LXD
1

Hi,

here’s what’s happening, I have a mass, lxd etc working. furthermore I have deployed charmed k8s, now to pass through GPUs I use the following:

$ lxc config device add hot-emu gpu0 gpu gputype=physical pci=0000:04:00.0
Device gpu0 added to hot-emu
:~$ lxc start hot-emu

the above just hangs indefinitely, hot-emu is a k8s worker and also I cannot start this worker even with the MAAS UI

I feel like there a juju way of doing this like

  juju set-constraints kubernetes-worker device=RTX2080=[1, ]|nvidia/gpu [0, 3]

issues is that I dont know how to specify exactly which GPU

info from lxc gives the below

Card 1:
    NUMA node: 0
    Vendor: NVIDIA Corporation (10de)
    Product: TU104 [GeForce RTX 2080] (1e82)
    PCI address: 0000:04:00.0
    Driver: nvidia (570.86.16)
    DRM:
      ID: 1
      Card: card1 (226:1)
      Control: controlD65 (226:1)
      Render: renderD128 (226:128)
    NVIDIA information:
      Architecture: 7.5
      Brand: GeForce
      Model: NVIDIA GeForce RTX 2080
      CUDA Version: 12.8
      NVRM Version: 570.86.16
      UUID: GPU-da3ef596-0ddd-0386-658d-1e3ee20f780b
Card 2:
    NUMA node: 1
    Vendor: NVIDIA Corporation (10de)
    Product: TU104 [GeForce RTX 2080] (1e82)
    PCI address: 0000:82:00.0
    Driver: nvidia (570.86.16)
    DRM:
      ID: 2
      Card: card2 (226:2)
      Control: controlD66 (226:2)
      Render: renderD129 (226:129)
    NVIDIA information:
      Architecture: 7.5
      Brand: GeForce
      Model: NVIDIA GeForce RTX 2080
      CUDA Version: 12.8
      NVRM Version: 570.86.16
      UUID: GPU-70d6028f-8add-f256-4a5f-ec4507189d9a

thanks