NVIDIA device name getting changed after patching ESXI from 9.0.0 to 9.0.1

Products

• VMware vSphere ESXi

Issue/Introduction

• After ESXI upgrade to 9.0.1 GPU card shows a different name  instead of NVIDIA
• This causes a hardware incompatibility during VM migration
• While running nvidia-smi command, the cards are not getting detected, returning a blank output.

Environment

VMware vSphere ESXI 9.0.1

Cause

NVIDIA cards were getting detected with a general naming convention GA107GL and hence when we run NVIDIA-SMI command, the output was not returned due to the an invalid  PCIE device detection.

 

Resolution

• Newer drivers were released by NVIDIA (570.195.02 or above) which should be used with image  For NVIDIA product matrix, please refer : NVIDIA-Product Support Matrix
• Involve NVIDIA team, to validate if the NVIDIA license supports 580.95.02. Reference:NVIDIA-Product Support Matrix
• Download the zip bundle for NVIDIA 580.95.02 and use the same with the image for the patch. Reference :Managing the Lifecycle of Clusters with vSphere Lifecycle Manager Images at Cluster Level
• Initiate the patch from vCenter Lifecycle manager and  perform an inventory sync in the SDDC manager. Synchronize the VCF Operations Fleet Management Inventory for Site Protection and Disaster Recovery for VMware Cloud Foundation