VM fails after a second vGPU is assigned to it
Description
After a second vGPU is added to a VM and the VM is restarted, the VM fails. NVIDIA vGPU software supports up to a maximum of 16 vGPUs per VM on VMware vSphere Hypervisor (ESXi).
When this issue occurs, the following messages are written to the log file on the hypervisor host:
2021-09-27T17:11:42.303Z| vthread-2105551| | I005: vmiop_log: (0x0): Start restoring vGPU state ...
2021-09-27T17:11:43.465Z| vcpu-0| | E002: vmiop_log: (0x0): Deferred restore for RPCs cannot continue, since restore data was not saved
2021-09-27T17:11:43.465Z| vcpu-0| | E002: vmiop_log: (0x0): Deferred call for vmiopd_restore_rpc_data failed at un-stun!
2021-09-27T17:11:43.465Z| vcpu-0| | E002: vmiop_log: (0x0): Failed to complete restore for deferred functions.
2021-09-27T18:44:27.034Z| vthread-2105550| | E002: vmiop_log: (0x0): VGPU message 1 failed, guest VGX version is already initialized...
2021-09-27T18:44:27.034Z| vthread-2105550| | E002: vmiop_log: (0x0): VGPU message 1 failed, result code: 0x40
...
2021-09-27T18:44:35.359Z| vthread-2105550| | I005: vmiop_log: (0x0): Guest driver unloaded!
Workaround
To avoid this issue, create your VMs in EFI mode.
If you encounter this issue with a VM that was created in legacy BIOS mode, shut down and restart the VM or power off the VM and power it on again.
Status
Not an NVIDIA bug
Ref. #
3386681