EKS example does not work by default

[jameslamb]

Description

I believe the walk-through at https://docs.rapids.ai/deployment/stable/cloud/aws/eks/ requires some modifications.

I'll add more details tomorrow, but in short... following that example, without modification, I saw nvidia-driver-daemonset pods from the gpu-operator helm chart getting stuck in ImagePullBackOff, with an error like this:

Failed to pull image "nvcr.io/nvidia/driver:550.90.07-amzn2": rc error: code = NotFound desc = failed to pull and unpack image "nvcr.io/nvidia/driver:550.90.07-amzn2": failed to resolve reference "nvcr.io/nvidia/driver:550.90.07-amzn2": nvcr.io/nvidia/driver:550.90.07-amzn2: not found

Notes

(placeholder: will add more soon)

References

Some relevant references I consulted while debugging this

• https://github.com/eksctl-io/eksctl/issues/6452
• https://docs.aws.amazon.com/eks/latest/userguide/eks-optimized-ami.html#gpu-ami?icmpid=docs_console_unmapped
• https://eksctl.io/usage/gpu-support/
• https://docs.nvidia.com/datacenter/cloud-native/gpu-operator/latest/amazon-eks.html#eks-node-group-with-the-gpu-operator
• https://docs.nvidia.com/datacenter/cloud-native/gpu-operator/latest/amazon-eks.html#about-using-the-operator-with-amazon-eks

• 195

• https://github.com/NVIDIA/gpu-operator