nvidia-smi ignores the second NVIDIA vGPU device added to a Microsoft Windows Server 2016 VM
Description
After a second NVIDIA vGPU device is added to a Microsoft Windows Server 2016 VM, the device does not appear in the output from the nvidia-smi command. This issue occurs only if the VM is already running NVIDIA vGPU software for the existing NVIDIA vGPU device when the second device is added to the VM.
The nvidia-smi command cannot retrieve the guest driver version, license status, and accounting mode of the second NVIDIA vGPU device.
nvidia-smi vgpu --query
GPU 00000000:37:00.0
Active vGPUs : 1
vGPU ID : 3251695793
VM ID : 3575923
VM Name : SVR-Reg-W(P)-KuIn
vGPU Name : GRID V100D-32Q
vGPU Type : 185
vGPU UUID : 29097249-2359-11b2-8a5b-8e896866496b
Guest Driver Version : 580.88
License Status : Licensed
Accounting Mode : Disabled
...
GPU 00000000:86:00.0
Active vGPUs : 1
vGPU ID : 3251695797
VM ID : 3575923
VM Name : SVR-Reg-W(P)-KuIn
vGPU Name : GRID V100D-32Q
vGPU Type : 185
vGPU UUID : 2926dd83-2359-11b2-8b13-5f22f0f74801
Guest Driver Version : Not Available
License Status : N/A
Accounting Mode : N/A
Version
This issue affects only VMs that are running Microsoft Windows Server 2016 as a guest OS.
Workaround
To avoid this issue, configure the guest VM with both NVIDIA vGPU devices before installing the NVIDIA vGPU software graphics driver.
If you encounter this issue after the VM is configured, use one of the following workarounds:
- Reinstall the NVIDIA vGPU software graphics driver.
- Forcibly uninstall the Microsoft Basic Display Adapter and reboot the VM.
- Upgrade the guest OS on the VM to Microsoft Windows Server 2019.
Status
Not an NVIDIA bug
Ref. #
3562801