NVIDIA GPU (OSS drivers)

Enabling NVIDIA GPU support on Talos is bound by NVIDIA EULA. The Talos published NVIDIA OSS drivers are bound to a specific Talos release. The extensions versions also needs to be updated when upgrading Talos.

We will be using the following NVIDIA OSS system extensions:

• nvidia-open-gpu-kernel-modules
• nvidia-container-toolkit

Create the boot assets which includes the system extensions mentioned above (or create a custom installer and perform a machine upgrade if Talos is already installed).

Make sure the driver version matches for both the nvidia-open-gpu-kernel-modules and nvidia-container-toolkit extensions. The nvidia-open-gpu-kernel-modules extension is versioned as <nvidia-driver-version>-<talos-release-version> and the nvidia-container-toolkit extension is versioned as <nvidia-driver-version>-<nvidia-container-toolkit-version>.

Proprietary vs OSS Nvidia Driver Support

The NVIDIA Linux GPU Driver contains several kernel modules: nvidia.ko, nvidia-modeset.ko, nvidia-uvm.ko, nvidia-drm.ko, and nvidia-peermem.ko. Two “flavors” of these kernel modules are provided, and both are available for use within Talos:

• Proprietary, This is the flavor that NVIDIA has historically shipped.
• Open, i.e. source-published/OSS, kernel modules that are dual licensed MIT/GPLv2. With every driver release, the source code to the open kernel modules is published on https://github.com/NVIDIA/open-gpu-kernel-modules and a tarball is provided on https://download.nvidia.com/XFree86/.

The choice between Proprietary/OSS may be decided after referencing the Official NVIDIA announcement.

Enabling the NVIDIA OSS modules

Patch Talos machine configuration using the patch gpu-worker-patch.yaml:

machine:
  kernel:
    modules:
      - name: nvidia
      - name: nvidia_uvm
      - name: nvidia_drm
      - name: nvidia_modeset
  sysctls:
    net.core.bpf_jit_harden: 1

Now apply the patch to all Talos nodes in the cluster having NVIDIA GPU’s installed:

talosctl patch mc --patch @gpu-worker-patch.yaml

The NVIDIA modules should be loaded and the system extension should be installed.

This can be confirmed by running:

talosctl read /proc/modules

which should produce an output similar to below:

nvidia_uvm 1146880 - - Live 0xffffffffc2733000 (PO)
nvidia_drm 69632 - - Live 0xffffffffc2721000 (PO)
nvidia_modeset 1142784 - - Live 0xffffffffc25ea000 (PO)
nvidia 39047168 - - Live 0xffffffffc00ac000 (PO)

talosctl get extensions

which should produce an output similar to below:

NODE           NAMESPACE   TYPE              ID                                                                           VERSION   NAME                             VERSION
172.31.41.27   runtime     ExtensionStatus   000.ghcr.io-siderolabs-nvidia-container-toolkit-515.65.01-v1.10.0            1         nvidia-container-toolkit         515.65.01-v1.10.0
172.31.41.27   runtime     ExtensionStatus   000.ghcr.io-siderolabs-nvidia-open-gpu-kernel-modules-515.65.01-v1.2.0       1         nvidia-open-gpu-kernel-modules   515.65.01-v1.2.0

talosctl read /proc/driver/nvidia/version

which should produce an output similar to below:

NVRM version: NVIDIA UNIX x86_64 Kernel Module  515.65.01  Wed Mar 16 11:24:05 UTC 2022
GCC version:  gcc version 12.2.0 (GCC)

Deploying NVIDIA device plugin

First we need to create the RuntimeClass

Apply the following manifest to create a runtime class that uses the extension:

---
apiVersion: node.k8s.io/v1
kind: RuntimeClass
metadata:
  name: nvidia
handler: nvidia

Install the NVIDIA device plugin:

helm repo add nvdp https://nvidia.github.io/k8s-device-plugin
helm repo update
helm install nvidia-device-plugin nvdp/nvidia-device-plugin --version=0.13.0 --set=runtimeClassName=nvidia

(Optional) Setting the default runtime class as nvidia

Do note that this will set the default runtime class to nvidia for all pods scheduled on the node.

Create a patch yaml nvidia-default-runtimeclass.yaml to update the machine config similar to below:

- op: add
  path: /machine/files
  value:
    - content: |
        [plugins]
          [plugins."io.containerd.cri.v1.runtime"]
            [plugins."io.containerd.cri.v1.runtime".containerd]
              default_runtime_name = "nvidia"        
      path: /etc/cri/conf.d/20-customization.part
      op: create

Now apply the patch to all Talos nodes in the cluster having NVIDIA GPU’s installed:

talosctl patch mc --patch @nvidia-default-runtimeclass.yaml

Testing the runtime class

Note the spec.runtimeClassName being explicitly set to nvidia in the pod spec.

Run the following command to test the runtime class:

kubectl run \
  nvidia-test \
  --restart=Never \
  -ti --rm \
  --image nvcr.io/nvidia/cuda:12.5.0-base-ubuntu22.04 \
  --overrides '{"spec": {"runtimeClassName": "nvidia"}}' \
  nvidia-smi