Use Mirrored Kale Python Image

To be able to run HP tunning trials with Katib in an air gapped environment, you have to configure Kale to use the mirrored kale-py38 image. This guide will walk you through enabling the corresponding PodDefault by default.

Procedure

  1. Go to your GitOps repository, inside your rok-tools management environment:

    root@rok-tools:~# cd ~/ops/deployments
  2. Edit kubeflow/manifests/apps/jupyter/jupyter-web-app/upstream/overlays/deploy/patches/config-map.yaml and extend the configurations list with the kale-python-image label:

    configurations: value: - access-rok - access-ml-pipeline - kale-python-image # <-- Add this line
  3. Commit changes:

    root@rok-tools:~/ops/deployments# git commit -am "Enable PodDefault for custom Kale Python image"
  4. Apply changes:

    root@rok-tools:~/ops/deployments# rok-deploy --apply \ > kubeflow/manifests/apps/jupyter/jupyter-web-app/upstream/overlays/deploy

Summary

You have successfully configured Kale to use the mirrored Kale Python image.

What’s Next

You may now go back and proceed with deploying Kubeflow.