Skip to content

GPU Passtrough support via kubevirt #96

@archenroot

Description

@archenroot

First of all thank you for this quite robust k8s config.
Feature Request
My Gentoo based workstation system has 2 gpu cards, 1 nvidia 1030 used for display and another nvidia Tesla (Architecture: Pascal
Variant: GP100-885-A1
Cuda cores: 2 x 3072). I would like to be able to bind this nvidia tesla to be available on the worker nodes, so I can deploy tensorflow models and use it on the cluster.

What should the feature do:
There are 2 projects which focus on gpu support in k8s nodes:
https://kubevirt.io/tag/passthrough
https://github.com/NVIDIA/kubevirt-gpu-device-plugin

I would like to use those for my aproach. Also note I use libvirt provider by default.

What would be solved through this feature:
GPU support on kube cluster.

Does this have an impact on existing features:
I don't think so, its an extending option to other features.

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions