Kubernetes

This article shows you how to start JupyterLab with the Roboweb extension on Kubernetes using our public docker image

• us-west1-docker.pkg.dev/roboweb-images/roboweb-images/kubeflow/jupyterlab:latest

  • Note the Docker image is publicly available but to browse the registry you must sign in with a Google account.

This docker image contains JupyterLab with the Roboweb extension already installed. It is based on Kubeflow’s JupyterLab image and thus should be deployable inside Kubeflow.

If you don’t have Kubeflow, you can use the minimal Kubernetes manifest in jlewi/roboweb-docs/k8s/manifests.

Clone the jlewi/roboweb-docs repository

git clone https://github.com/jlewi/roboweb-docs.git

Note As is this manifest isn’t intended for doing meaningful work inside JupyterLab. In particular, it doesn’t use persistent disk to store your notebooks so any work will be lost when the pod restarts. For a more robust notebook deployment we recommend

• Using Kubeflow Notebooks
• Customizing the manifests to meet your needs

Deploying JupyterLab on Kubernetes

To deploy it run

cd roboweb-docs
kubectl create namespace roboweb
kustomize build k8s/manifests | kubectl apply -f

This will deploy the notebook in the namespace roboweb.

To use a different namespace edit the k8s/manifests/kustomization.yaml file.

Accessing It

To access JupyterLab port-forward to the pod

kubectl -n roboweb port-forward notebook-0 8888:8888

Then open a browser to http://localhost:8888 to access JupyterLab.

Using Roboweb in JupyterLab

Refer to the directions in the quickstart for how to use Roboweb in JupyterLab.

You can skip the steps to install Roboweb since the docker image already has it installed.