Rancher Cluster Management: Inject Helm Charts on Provisioning Time

SUSE Rancher Prime
1

Hello,

I’m provisioning a single-node K3s cluster via Rancher on a machine equipped with NVIDIA GPUs.

To expose the GPU to the cluster, I need to deploy the nvdp/nvidia-device-plugin Helm chart. While this can be done manually post-provisioning using Rancher’s Apps feature, it’s cumbersome and not scalable for larger environments.

I noticed there’s an option to apply additional Kubernetes manifests during cluster provisioning, but not Helm charts.

Screen Shot 2025-06-30 at 10.03.49
Screen Shot 2025-06-30 at 10.03.49918×346 23.4 KB

What’s the recommended or best-practice approach to automatically deploy the NVIDIA device plugin (or other system-level Helm charts) when provisioning a cluster via Rancher?

Any guidance would be appreciated!

Thanks in advance!

Sam

2

Use fleet and it’s gitops features?

3

As stated above, Fleet would be the most integrated solution. Another option is to use the HelmChart controller. If using RKE2, it comes pre-installed and if not, you should be able to both install the controller and deploy the HelmChart CRs through the additional manifest fields.