For RHODS we plan to put a taint on GPU worker nodes so that we can restrict their use to only workloads that actually need GPUs.

Before we can apply the taint, we need to modify the JupyterHub profiles code so that, if a user requests GPUs for their notebook pod, they get the necessary toleration to land on a GPU worker node.

As such, we should add support for a config section in the Jupyterhub Single User Profiles ConfigMap to specify information related to GPUs. The config section would look similar to 

  gpu:
  - type: ...
    taint: ....

From vpavlin@redhat.com:

You will want to add the code here: https://github.com/opendatahub-io/jupyterhub-singleuser-profiles/blob/master/jupyterhub_singleuser_profiles/profiles.py#L206

Probably refactor this https://github.com/opendatahub-io/jupyterhub-singleuser-profiles/blob/master/jupyterhub_singleuser_profiles/profiles.py#L294-L300 out to a separate method and just call that from the apply_gpu_config when gpu_count == 0