VMs with vGPU not powering on after upgrade to ESXi 6.5u1

[jseide86]

Hello

I just upgraded my 6.5d hosts to 6.5 update 1 (5969303) and the NVIDIA GRID Host Driver 384.73 (4598673) on my Cisco UCS C240-M4L servers with Nvidia Tesla M10 cards. My vms, which use the "M10-1B" profile, refuses to start.

I'm getting the symptoms described in this KB: “Could not initialize plugin '/usr/lib64/vmware/plugin/libnvidia-vgx.so' for vGPU "profile_name"” er... but the resolution does nothing for me. The "nvidia-smi" command shows normal operating status, and the host seems correctly configured (Shared Direct mode for the GPUs).

Any tips? I've ended up empty handed after googling this issue.

[matteoabrile]

Hello,

I sove put in host -> Configure -> Graphics -> Shared Direct

Try with this change.

Bye

M.

[jseide86]

Hi,

My hosts are already configued with this option, and is what's described in the KB i linked in my original post. This does not resolve anything.

[LucaMeglioli]

hi ,

has anyone found a solution?

we've just intsalled a vmware patch for vsan, and after that we are able to power on 6 VM with 2q profile (6 GB of Memory used for each GPU , with 2 more GB free)

but when starting the other VM we got the error.

if i power down one of the 6 VM i'm able to power on another one... seems that the system cannot handle correctly the last 2 GB of memory of the GPU (but we have not the classic "insufficient graphic resource " error...)