vGPU VM crashes after vMotion completion

Products

• VMware vSphere ESXi

Issue/Introduction

Following successful vMotion, the vGPU (virtual Graphics Processing Unit) VM experiences an unexpected crash.

The VM's vmware.log contains the following information:

<YYYY-MM-DD>T<HH>:01:56.464Z In(05) vcpu-0 - MigrateSetState: Transitioning from state MIGRATE_FROM_VMX_FINISHED (12) to MIGRATE_TO_VMX_NONE (0).

...

<YYYY-MM-DD>T<HH>:02:05.775Z Wa(03) vcpu-3 - WinBSOD: Synthetic MSR[0x40000100] 0x7e

The /var/run/log/vmkernel.log contains the following information:

<YYYY-MM-DD>T<HH>:01:58.967Z In(182) vmkernel: cpu19:######2)NVRM: Xid (PCI:0000:##:00): 43, channel 0x00000##2

...

<YYYY-MM-DD>T<HH>::02:01.016Z In(182) vmkernel: cpu19:######2)NVRM: Xid (PCI:0000:##:00): 43, channel 0x01000##5

The /var/run/log/syslog contains the following information:

<YYYY-MM-DD>T<HH>:01:58.969Z In(30) nvidia-xid-logd[######7]: XID 43 detected on device index: 0, initiating dump...

...

<YYYY-MM-DD>T<HH>:02:01.017Z In(30) nvidia-xid-logd[######7]: XID 43 detected on device index: 0, initiating dump...

Environment

VMware vSphere ESXi 8.0.x

Resolution

Engage the Guest OS vendor to investigate the cause of the VM crash. (Refer to Troubleshooting a Virtual Machine that has stopped responding for details)

Engage the GPU vendor to investigate the XID event.

Ensure that the latest ESXi host VIBs and Guest OS drivers are used for the GPUs. (Refer to DirectPath GPUs and Accelerators for such information)

 

Additional Information

Refer to Analyzing Xid Errors with the Xid Catalog for XID information.

Refer to vGPU VM becomes unresponsive with vmotion for additional information.