SLES-based RKE2 support

[hase1128]

Current NVIDIA GPU Operator does not support SUSE Linux Enterprise Server(SLES)-based Rancher Kubernetes Engine 2(RKE2). It seems that no container images of "NVIDIA GPU driver and container toolkits" for SLES are actually provided.
It supports only Ubuntu, RHEL and RHCOS(https://docs.nvidia.com/datacenter/cloud-native/gpu-operator/latest/platform-support.html#supported-operating-systems-and-kubernetes-platforms)

We are planning to provide our customers with infrastructure services that utilize GPUs on SLES-based RKE2. It would be useful to be able to deploy and manage the necessary components, including drivers, using the GPU Operator.
Therefore, could you please support SLES-based RKE2 for NVIDIA GPU Operator?

[shan100github]

I was using RKE2 with Ubuntu OS, and it worked fine even before 2022.
I don't know anything specific about SLES.

Also, I came across https://docs.rke2.io/advanced#deploy-nvidia-operator

probably https://docs.nvidia.com/datacenter/cloud-native/gpu-operator/latest/platform-support.html#supported-operating-systems-and-kubernetes-platforms doc to be updated 🤔

[hase1128]

Some container images deployed and managed by the GPU Operator are OS-dependent.
For example, the driver container image is not provided for SLES.

[alex-isv]

You can just build your own container driver and host it on your local repo or install cuda-drivers on workers and disable a driver from the Operator helm during deployment if you don't have too many worker nodes. I tested it many times for years and it's working great. Here is an example - Deploying Nvidia GPU-Operator in SLES based RKE2 cluster

[hase1128]

Yes, currently we have to manually install the NVIDIA GPU drivers directly on the host.
(It should not be installed by the GPU Operator.)
However, this is inconvenient because it increases operational costs for users. Ideally, the GPU Operator should be able to manage the lifecycle of all necessary components.

[francisguillier]

Hi,
I’ve added the request to our backlog, and we’ll prioritize it as we see increased demand.
Changing the label for this issue with 'feature' as RFE for tracking.