VMware Cloud Community
PHGustavsson
Contributor
Contributor

BSOD on one VM with DirectPath I/O enabled (vDGA - GRID K2)

We have two Dell R720 with one Nvidia GRID K2 card in each server. Running ESXi 5.5.

Two XenApp (v. 6.5 on Win2k8r2) VM's each with vDGA enabled.

Nvidia drivers installed on the hosts and on the VM's.

Like this:
ESX01 - XenApp01 and XenApp02
ESX02 - XenApp03 and XenApp04

XenApp04 is crashing alot. Always caused by nvlddmkm.sys:
TESTVM BSOD.png

If I switch the GPU assigned to XenApp04 with the one assigned to XenApp03 then XenApp03 starts to crash.

I installed a new VM with Win2k8r2 on ESX02 and assigned the GPU from XenApp04 to that VM and it crash randomly when I run GPU-Z and/or benchmark tools from RDP.

As you can imaging XenApp04 is not in production anymore and has been shut down for a few days now.

When I today tried to start XenApp04 with the GPU assigned it boot loops with bluescreens in the Windows boot, this time caused by dxgmms1.sys:
XENAPP02 BSOD.png

Is that because other VM's on ESX02 are using the GPU in vSGA (the drivers are installed for vSGA on the host as I mentioned)? The GPU was assigned to XenApp04 the entire time it was shut down.

Or is the K2 card on ESX02 faulty?

Anything else I can test and/or do?

I will contact Dell and let them troubleshoot it but just want to check here first before doing so.

Regards
P-H, Sweden

0 Kudos
1 Reply
PHGustavsson
Contributor
Contributor

Nobody? Smiley Sad

0 Kudos