aihub-dag/docs/source/recipes/using-elyra-with-kubeflow-notebook-server.md

Using Elyra with the Kubeflow Notebook Server

In this example we will show how to launch Elyra using Kubeflow's Notebook Server.

Requirements

• A Kubeflow Cluster

  • A standard installation of Kubeflow will include the Notebook Server as part of the application stack
  • Installation Documentation

• Select the desired Elyra container image version from the elyra/kf-notebook repositories on Docker Hub or quay.io. Do NOT use other images, such as the ones in elyra/elyra.

OR

Create a custom Elyra container image following the instructions in this directory.

Launching Elyra in the Kubeflow Notebook Server

1. In the default Kubeflow welcome page, in the left side menu, click on Notebook Servers
   
   

1. Click on NEW SERVER
   
   

1. Choose a name for your notebook server, and under Image check the box labeled Custom Image.
   
   

1. In Kubeflow version 1.3 (and later) choose jupyterlab as image type.

2. As Custom Image enter elyra/kf-notebook:<ELYRA_VERSION>, replacing <ELYRA_VERSION> with the desired image tag, e.g. 2.1.0.

To pull an image from quay.io prefix the image name with quay.io, e.g. quay.io/kf-notebook:2.1.0.

1. Customize the resources as required. We recommend at least 1 CPU and 1Gi of memory.

2. Your notebook server with Elyra should begin the provisioning process now. When complete, you'll notice a green light and arrow to the left of your notebook. Click on Connect to launch Elyra.

Next step:

Creating a runtime metadata configuration

Additional Resources and Documentation

Official Kubeflow Notebook Server Documentation - Creating a Jupyter Notebook Server