VxRail: Xorg Service Fails To Start With Pci-Passthrough After Installing NVIDIA GPU VIB
Summary: VxRail: After installing the NVIDIA Graphics Processing Unit (GPU) VIB file, xorg service fails to start when using pci-passthrough.
Symptoms
Identified Symptoms
The xorg service fails to start when using pci-passthrough after installing the NVIDIA GPU VIB file.
The output of the command to check the xorg stats indicates that the service is not running.
/etc/init.d/xorg status
The output shows:
Xorg is not running
Also, the command "nvidia-smi" fails.
nvidia-smi
Running the command to list the PCI devices shows the ModuleName as "pciPassthru" instead of the expected "Nvidia" for the NVIDIA devices.
esxcli hardware pci list -c 0x0300 -m 0xf
The output includes:
ModuleName: pciPassthru
Instead of the expected:
ModuleName: nvidia
Cause
Resolution
Resolving the xorg Service Failure
To resolve the issue with the xorg service failing to start when using pci-passthrough, follow these steps:
- Navigate to the vSphere web client
- Select the host --> Hardware --> PCI Devices --> Click edit (in vSphere client, you find the configure passthrough button in the upper right corner)
- Change the status of all the NVIDIA devices to "Unavailable" (or at least one of them)
- Navigate then to host --> Hardware --> Graphics
- Confirm that the graphics type is "Shared direct" and not just "Shared"
- After that select the host --> configure --> system --> security profile --> services --> xorg
- Click edit, and select Start and stop with host and then click ok
- Put the node into maintenance mode and make sure all the VMs are migrated
- Reboot the node
The xorg service should be running after these steps. You can verify this by running the following commands:
/etc/init.d/xorg status nvidia-smi
These commands should show that the xorg service is running and the NVIDIA GPU is recognized, respectively.