Linux VM hangs after vGPU migration to a host running a newer vGPU manager version

Description

When a Linux VM configured with a Tesla V100 or Tesla T4 vGPU is migrated from a host that is running a vGPU manager 11 release before 11.6 to a host that is running a vGPU manager 13 release, the VM hangs. After the migration, the destination host and VM become unstable. When this issue occurs, XID error 31 is written to the log files on the destination hypervisor host.

Version

This issue affects migration from a host that is running a vGPU manager 11 release before 11.6 to a host that is running a vGPU manager 13 release.

Workaround

If the VM is configured with a Tesla T4 vGPU, perform the following sequence of steps before attempting the migration:

Upgrade the host that is running a vGPU manager 11 release to release 11.6 or a later vGPU manager 11 release.
Disconnect any remoting tool that is using NVENC.

Note:

You cannot use this workaround for a VM that is configured with a Tesla V100 vGPU.

Status

Open

Ref. #

200691445