GPU Gone from Truenas Scale

Hello. Hope anyone here is able to help me.

I got a a TrueNas server running on the following hardware:

i7 8700k
48gb ram
Nvidia 1660ti
msi z370 gaming m5

Truenas version: ElectricEel 24.10

Today i brought in the 1660ti gpu to do some vm’s and stuff on it.

When i plugged in the gpu and booted into TrueNas i was able to see the gpu in the devices section when trying to create a new VM. However it showed up as " unknown 02:000 " something. And upon creating the VM and booting it up this error showed up:

[EFAULT] internal error: qemu unexpectedly closed the monitor: 2024-11-17T22:13:14.945288Z qemu-system-x86_64: -device {“driver”:“vfio-pci”,“host”:“0000:02:00.0”,“id”:“hostdev0”,“bus”:“pci.0”,“addr”:“0x7”}: vfio 0000:02:00.0: group 2 is not viable Please ensure all devices within the iommu_group are bound to their vfio bus driver.

I did lots of googling, reading forums etc and tried the following:

Checked my bios settings, V-td is enabled. intel virtualiztion is enabled - Onboard gpu is set to PEG ( onboard gpu ) and multimonitor set to disabled.

I tried updating the grub file with

  1. intel_iommu=on iommu=pt
  2. pcie_acs_override=downstream
  3. nomodeset i915.modeset=0

In all different ways, and ofcourse remembered to do " update-grub "

None of which works. I want to note also that after i did the " intel_iommu=on " updated grub and rebooted the GPU since hasn’t showed up not even once when trying to create a new VM.

Last but not least. the GPU shows up fine when i do " lspci " - and the nvidia-smi command reports back with 1660ti - Its looks like this:


lspci looks like this:

For some reason i am able to see the GPU in the apps/docker section and thats it.

However, when trying to enable it for an app it gives me this:

[EFAULT] Failed to render compose templates: base_v1_1_7.utils.TemplateException: Expected [uuid] to be set for GPU inslot [0000:01:00.0] in [nvidia_gpu_selection]

Hope anyone is able to help

Haven’t used your board before but this sounds like you’ve disabled the iGPU which tracks with your lspci output. You need to set your graphics choice to “onboard first” or possibly “onboard only” and maybe “multimonitor enabled”

What you want to see is both your NVIDIA and Intel GPU show up as VGA devices in the lspci output. As it is you won’t be able to isolate the 1660 for VM use because it’s your only card.

This is a separate Docker issue resolved by the command in this thread:

1 Like