2

I am struggling to add a GPU to my GKE Kubeflow cluster. The documentation still references kfctl and some old set-up parameters. (To be precise, I added a T4 GPU to the GKE cluster successfully, but my notebook server fails to start).

Has anyone managed to successfully add GPUs (btw I only have 1 in my dev project) and launch a notebook server with GPUs? If so, how did you do it and what NVIDIA GPU are you using?

OlgaPp
  • 180
  • 11
  • Are you trying to add a GPU node pool to your cluster or a GPU to a node pool? Because, as per the [limitations on using GPUs on GKE](https://cloud.google.com/kubernetes-engine/docs/how-to/gpus#limitations), you cannot add GPUs to existing node pools. You can only [add a GPU node pool to your cluster](https://cloud.google.com/kubernetes-engine/docs/how-to/gpus#gpu_pool). – Ksign Nov 11 '20 at 16:08
  • I created a separate node pool with GPU. So, under my KF cluster, there are two node pools: default and a gpu one. – OlgaPp Nov 12 '20 at 10:48
  • I am able to run a pipeline on GPU (when triggering the run out of SDK), but am unable to launch a notebook – OlgaPp Nov 12 '20 at 10:48

0 Answers0