GRID: NVIDIA Virtual GPU Manager on ESXi host

From Define Wiki
Revision as of 13:35, 1 April 2020 by Danrogers (talk | contribs) (Completed page.)
Jump to navigation Jump to search

Nvidia: Landing Page : GRID 10.1 VMware stack installation and testing

Install NVIDIA Virtual GPU Manager on ESXi Host

  1. Get version number of latest 'Driver Package' from | NVIDIA Virtual GPU Software Supported Products
  2. Download NVIDIA Virtual GPU Manager | NVIDIA Licensing Software Downloads
    E.g. For NVIDIA vGPU for vSphere 6.7
    \\10.0.0.222\software\NVIDIA\GRID\10.1\VMware vSphere\NVIDIA-GRID-vSphere-6.7-440.53-440.56-442.06.zip
  3. Extract the archive
  4. Ensure SSH is enabled on the ESXi host
  5. Copy the 'vib' file via SFTP to the ESXi host to the system root '/'
  6. Put ESXi host into maintenance mode
  7. SSH onto the ESXi host
  8. Run the following command:
    [root@localhost:~] esxcli software vib install -v /NVIDIA-VMware_ESXi_6.7_Host_Driver-440.53-1OEM.670.0.0.8169922.x86_64.vib
    IMAGE REQUIRED Below is the 'operation finished successfully' message.
    Installation Result
    Message: Operation finished successfully.
    Reboot Required: false
    VIBs Installed: NVIDIA_bootbank_NVIDIA-VMware_ESXi_6.7_Host_Driver_440.53-1OEM.670.0.0.8169922
    VIBs Removed:
    VIBs Skipped:
  9. Restart ESXi host
    esxcli system shutdown reboot -r reason_for_shutdown
  10. 'Exit maintenance mode' on the ESXi host
  11. Enable SSH on ESXi host
  12. SSH onto ESXi host
  13. Run the following command to validate a NVIDIA module is loaded into the kernel:
    [root@localhost:~] vmkload_mod -l | grep nvidia
    IMAGE REQUIRED Expected output:
    [root@localhost:~] vmkload_mod -l | grep nvidia
    nvidia 35 19360
  14. Run the following command to validate the 'NVIDIA Virtual GPU Manager' is running as expected:
  15. [root@localhost:~] nvidia-smi
    IMAGE REQUIRED Expected output, you can see a 'Quadro RTX 6000' and 24,575 MiB of memory is present:
    Mon Mar 23 10:49:18 2020
    +-----------------------------------------------------------------------------+
    | NVIDIA-SMI 440.53 Driver Version: 440.53 CUDA Version: N/A |
    |-------------------------------+----------------------+----------------------+
    | GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
    | Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
    |===============================+======================+======================|
    | 0 Quadro RTX 6000 On | 00000000:89:00.0 Off | Off |
    | 33% 30C P8 28W / 260W | 179MiB / 24575MiB | 0% Default |
    +-------------------------------+----------------------+----------------------+
    +-----------------------------------------------------------------------------+
    | Processes: GPU Memory |
    | GPU PID Type Process name Usage |
    |=============================================================================|
    | 0 2100034 G Xorg 5MiB |
    +-----------------------------------------------------------------------------+

House Keeping

  1. Delete the vib file from the system root
    [root@localhost:~] rm -f /NVIDIA-VMware_ESXi_6.7_Host_Driver-440.53-1OEM.670.0.0.8169922.x86_64.vib