NVIDIA/k8s-device-plugin

Docker image tag v0.9.0-ubuntu20.04

yuliyan-valchev-ft opened this issue · 5 comments

yuliyan-valchev-ft commented

1. Quick Debug Information

  • OS/Version(e.g. RHEL8.6, Ubuntu22.04):
  • Kernel Version:
  • Container Runtime Type/Containerd
  • K8s Flavor/EKS 1.30

2. Issue or feature description

Hello,

We have created a new EKS 1.30 cluster with a gpu instance node group. On the cluster installed from AWS we get an nvidia-device-plugin daemonset and got notified that it is using an EOL Ububntu 16.04 base image.

Which lead me to look at the images for nvcr.io/nvidia/k8s-device-plugin:v0.9.0 and found that the there might be an error with the tag v0.9.0-ubuntu20.04, being actually on the ubuntu16.04 base

Here is a quick check:

docker run -it --platform=linux/amd64 --entrypoint bash nvcr.io/nvidia/k8s-device-plugin:v0.9.0-ubuntu20.04 

root@76e879e636b3:/# cat /etc/os-release
NAME="Ubuntu"
VERSION="16.04.7 LTS (Xenial Xerus)"
ID=ubuntu
ID_LIKE=debian
PRETTY_NAME="Ubuntu 16.04.7 LTS"
VERSION_ID="16.04"
HOME_URL="http://www.ubuntu.com/"
SUPPORT_URL="http://help.ubuntu.com/"
BUG_REPORT_URL="http://bugs.launchpad.net/ubuntu/"
VERSION_CODENAME=xenial
UBUNTU_CODENAME=xenial

Maybe the tag needs to be deleted as it is all the same image

image
elezar commented

@yuliyan-valchev-ft it is not clear why you are using such an old devic eplugin image. Are you able to specify the device plugin version? We have just released v0.16.2 and this should be used if possible.

elezar commented

It seems as if the EKS documentation here is out of date.

Screenshot 2024-08-08 at 16 32 25

Our recommended mechanism for installing and managing the plugin is Helm. The instructions from the README should be ablicable to EKS too.

vara-bonthu commented

We are working on updating these docs. Will update the issue once it's resolved.