Submit Kubernetes Resources
- Next steps

Submit Kubernetes Resources

Submitting Kubernetes resources from a Jupyter notebook

Kubeflow assigns the default-editor service account to the Jupyter notebook Pods. This service account is bound to the kubeflow-edit ClusterRole, which has namespace-scoped permissions to the many Kubernetes resources including:

Pods
Deployments
Services
Jobs
TFJobs
PyTorchJobs

You can get a full list of these permissions using:

kubectl describe clusterrole kubeflow-edit

You can therefore create the above Kubernetes resources directly from your Jupyter notebook in Kubeflow. The Kubernetes kubectl command-line tool is pre-installed in the notebook.

Run the following command in a Jupyter notebook cell to create Kubernetes resources:

!kubectl create -f myspec.yaml

The myspec.yaml file should describe one of the above Kubernetes resources. For information about the format of the YAML file, see the Kubernetes object guide.

Next steps

See the guide to setting up your Jupyter notebooks in Kubeflow.
Explore the components of Kubeflow, including custom Kubernetes resources.

Last modified 04.03.2021: Remove outdated banners, move GCP instructions to GKE folder (#2525) (44ba9bb0)