Shengbo Song
Shengbo Song
The gpu-admission doesn't has the view of your `nvidia.com/gpu` pods, so it think the card 0 is the best fit for your pod, but gpu-manager has the real card usage,...
> High load average will show up when gpu-manager is deployed with prometheus. The load average values will go straight from 0 to 5000. There are about number-of-active-pods times that...
This seems a bug, it'll be fixed ASAP.
You request 100 core, that means you want a full gpu card, there is no limit when you did this.
Any kind of card after Kepler architecture.
Please provide the vcuda-controller log. About how to dump log, please see the FAQ of [gpu-manager](https://github.com/tkestack/gpu-manager)
Did you reboot your machine after upgrading your driver?
What's the version of your deployed gpu-manager？We have fixed this in our latest commit `808ff8c29a361f04499ff62242cd56e4f93089f6`
Upgrade to v1.1.2
> > Upgrade to v1.1.2 > > i use this version, Problem still exists Is there any log show `Read from`