bottlerocket-os / bottlerocket-os/bottlerocket

Add nvidia.com/gpu.present=true as a default label in NVIDIA variants

Open
#4,756 1 comment 2 reactions 0 assignees View on GitHub
area/kubernetes status/needs-proposal type/enhancement
Dominant language
Rust
Stars
9.7k
Forks
586
Avg merge
1d 11h
Merged PRs (30d)
11

Description

**What I'd like:**
The GPU Operator will automatically apply this label to hosts in the cluster that actually have GPUs (it leverages on NFD to detect the GPUs). Bottlerocket NVIDIA variants should apply this label to the nodes they run on, as we know that the variant will only run when NVIDIA GPUs are present. This label is required to run the NVIDIA DRA Plugin, as the official helm chart provided by NVIDIA uses the `nvidia.com/gpu.present=true` label as an affinity rule.

The EKS Optimized AMIs recently applied this change too: https://github.com/awslabs/amazon-eks-ami/pull/2607

**Any alternatives you've considered:**
The administrators can apply the label to the nodes, but that's too manual as they will have to do it whenever a node joins a cluster.

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.