The goal of this action was to upgrade a cluster with nvidia graphics card from 6.7 to 7.0.3. Before the upgrade everything worked fine. You where able to live vmotion vm:s with vgpu to another host.
Everything worked from vm side, but was not able to vmotion live vm:s between hosts.
- Upgrade esxi
- Upgrade nvidia driver
- Start vm:s
- Vmotion failed between hosts
This change must be applied on every host in the cluster that you want to live vmotion machines in between.
How to see if ECC is enabled or disabled.
nvidia-smi -q | grep "Ecc Mode" -A2
If your Current config is enabled you need to disable ECC, and you do this with this command:
nvidia-smi -e 0
And reboot is required!
And do not forget 1 host like I did! 🙂
Leave a Reply