Hi, we have come across an error where we are unable to get the density we should be able to on our Nvidia K1 cards. Specifically, with allocating 512MB of Video RAM to desktops, we are not approaching 64 VM's per GPU, as we should be able to. I have a case open with VMware and they have suggested that I should update my Nvidia drivers on ESXi host. Currently running 304.76-1OEM.510.0.0.802205, and I found the version 319.65-1OEM.510.0.0.802205 on the Nvidia site. If I try to run esxcli software vib install -v NVIDIA-VMware_ESXi_5.1_Host_Driver_319.65-1OEM.510.0.0.802205.vib then I get the error:
[InstallationError]
Error in running rm /tardisks/NVIDIA_V.v00:
Return code: 1
Output: rm: can't remove '/tardisks/NVIDIA_V.v00': Device or resource busy
If I try to run:
esxcli software vib remove --vibname=NVIDIA-VMware_ESXi_5.1_Host_Driver
then I get the same error.
All VM's are stopped and my host is in maintenance mode. How do I uninstall 304.76 driver so I can install 319.65 driver? Is there a service I need to stop first or something?
Here is the full procedure:
To update the NVIDIA GPU VIB, you must uninstall the currently installed VIB and install the new VIB, as described above.To uninstall the currently installed VIB:
/etc/init.d/xorg stop
vmkload_mod -u nvidia
esxcli software vib remove -n NVIDIA-VMware
Try stopping the xorg service:
/etc/init.d/xorg stop
Ok, that worked to allow me to install the new driver. However, I cannot access GPU resources on the host.
When I run 'nvidia-smi list' I get the message:
NVIDIA: API mismatch: the NVIDIA kernel module has version 304.76,
but this NVIDIA driver component has version 319.65. Please make
sure that the kernel module and all NVIDIA driver components have the same version.
Failed to initialize NVML: Unknown Error
The driver install said I didn't need to reboot the host, but should I reboot it anyway? Or is there something else I need to do to update the kernel module?
Also, I did run /etc/init.d/xorg start after the install and it said Xorg 0 through 7 started.
Hmm, but when I run xorg status, it says it's not started. I'm going to reboot the host and see if that resolves.
Did you download the right driver? There is one for vSphere 5.1 and another for vSphere 5.5
Here is the full procedure:
To update the NVIDIA GPU VIB, you must uninstall the currently installed VIB and install the new VIB, as described above.To uninstall the currently installed VIB:
/etc/init.d/xorg stop
vmkload_mod -u nvidia
esxcli software vib remove -n NVIDIA-VMware
I do have the right download, yes. However I did not run step 5, which could be my problem. I will try and report back. Thanks.
If you have bad consolidation ratios I would also recommend to update the vBios on the GPU, you need to contact you oem vendor to get it. (HP, IBM, Dell etc)