Hello community,
first of all, thank you for your last support and advice, it was completely helpful and I could solve the problem.
But I have another problem, and maybe you could help me once again. We’re using several DL385 Gen 11 Servers with L4 Cards and another Server with RTX 6000 cards. Everything works fine, setup was a little bit confusing, but at least easy to handle, after a few weeks of reading citrix and nvidia KBs, articles and threads.
Setup:
Fujitsu Primergy RX2540 M6 with 1 TB of RAM, 2x XEON Gold 6338 (32 physical, 32 logical cores on each of the CPUs), SSDs, and 2x Tesla RTX A6000 48GB PCIe. Actual BIOS Version on the Fujitsu Server.
OS is XenServer 8.4, nvidia vGPU Manager v19.1 for XenServer 8.4 installed and running (? I’ve checked that, but I’m not sure if it works correctly), and nvidia vWS Cloud Licensing.
Problem:
Here’s what im stucking on: When I’m setting up a new VM, which should be my goldimage for Citrix MCS and round about 10 Worker on the Fujitsu Server, the VM refuses to start with the error: The NVIDIA GPU is not configured for SR-IOV as expected.
I’m able to choose the correct profiles in XenCenter and both A6000 are also shown in XenCenter. If I choose a GPU Passtrough for the VM, it works. The VM runs. But if i choose a profile, like 8Q or 4A for example, the error occurs.
Here’s what I’ve done/checked yet:
· BIOS Settings of the Fujitsu server: ENABLED Above 4G Decoding, Intel Vt-D (IOMMU) ENABLED, SR-IOV ENABLED, Intel Virtualization ENABLED
· Checked if nvidia vGPU Software is installed with nvidia-smi command (shows both cards, but: “no processes are running” is shown?)
· Checked with command nvidia-smi vgpu: The cards are shown, but no profiles are shown like grid_a6000-6Q
· Compared the settings on the nvidia L4 Servers with the Fujitsu with nvidia-smi -q, they seem to be equal
· Reducing the RAM to 768 GB from 1 TB (I think I have read an article in the nvidia KB, that more than 768 shouldn’t be used)
Does anyone have an idea what could be wrong? Thank you very much for your help and best regards!
SOLUTION: used the nvidia Display Mode Selector Tool and changed the mode to **“**physical_display_disabled” on both cards. now the VM is starting as usual.