Quickstart Guide

Getting started with Kubeflow Notebooks


  1. Install Kubeflow by following Getting Started - Installing Kubeflow.
  2. Open the Kubeflow Central Dashboard in your browser.
  3. Click “Notebooks” in the left-hand panel.
  4. Click “New Server” to create a new notebook server.
  5. Specify the configs for your notebook server.
  6. Click “CONNECT” once the notebook has been provisioned

Detailed Steps

  1. Open the Kubeflow Central Dashboard in your browser.

  2. Select a Namespace:

    • Click the namespace dropdown to see the list of available namespaces.
    • Choose the namespace that corresponds to your Kubeflow Profile. (See the page on profiles and namespaces for more information)

    Selecting a Kubeflow namespace

  3. Click “Notebook Servers” in the left-hand panel:

    Opening notebooks from the Kubeflow UI

  4. Click “New Server” on the “Notebook Servers” page:

    The Kubeflow notebook servers page

  5. Enter a “Name” for your notebook server.

    • The name can include letters and numbers, but no spaces.
    • For example, my-first-notebook.

    Form for adding a Kubeflow notebook server

  6. Select a Docker “Image” for your notebook server

    • Custom image: If you select the custom option, you must specify a Docker image in the form registry/image:tag. (See the guide on container images.)
    • Standard image: Click the “Image” dropdown menu to see the list of available images. (You can choose from the list configured by your Kubeflow administrator)
  7. Specify the amount of “CPU” that your notebook server will request.

  8. Specify the amount of “RAM” that your notebook server will request.

  9. Specify a “workspace volume” to be mounted as a PVC Volume on your home folder.

  10. (Optional) Specify one or more “data volumes” to be mounted as a PVC Volumes.

  11. (Optional) Specify one or more additional “configurations”

    • These correspond to PodDefault resources which exist in your profile namespace.
    • Kubeflow matches the labels in the “configurations” field against the properties specified in the PodDefault manifest.
    • For example, select the label add-gcp-secret in the “configurations” field to match to a PodDefault manifest containing the following configuration:
    apiVersion: kubeflow.org/v1alpha1
    kind: PodDefault
      name: add-gcp-secret
      namespace: MY_PROFILE_NAMESPACE
        add-gcp-secret: "true"
     desc: "add gcp credential"
     - name: secret-volume
       mountPath: /secret/gcp
     - name: secret-volume
        secretName: gcp-secret
  12. (Optional) Specify any “GPUs” that your notebook server will request.

    • Kubeflow uses “limits” in Pod requests to provision GPUs onto the notebook Pods (Details about scheduling GPUs can be found in the Kubernetes Documentation.)
  13. (Optional) Specify the setting for “enable shared memory”.

    • Some libraries like PyTorch use shared memory for multiprocessing.
    • Currently, there is no implementation in Kubernetes to activate shared memory.
    • As a workaround, Kubeflow mounts an empty directory volume at /dev/shm.
  14. Click “LAUNCH” to create a new Notebook CRD with your specified settings.

    • You should see an entry for your new notebook server on the “Notebook Servers” page
    • There should be a spinning indicator in the “Status” column.
    • It can take a few minutes for kubernetes to provision the notebook server pod.
    • You can check the status of your Pod by hovering your mouse cursor over the icon in the “Status” column.
  15. Click “CONNECT” to view the web interface exposed by your notebook server.

    Opening notebooks from the Kubeflow UI

Next steps


Was this page helpful?