Connecting Jupyter/Kubeflow to Spark

I’m looking for some guidance on how to connect a Jupyter Notebook (the one shipped with Kubeflow) to Spark. Each are running on Kubernetes.

The Spark images on the docker-stacks repo don’t seem to work on Kubeflow/JupyterHub. The contains start, but Jupyter isn’t proxying the connections back to them correctly. The logs are all http 404s.

Any pointers appreciated.