This section gives the community a space to provide information on setting up, managing, and maintaining JupyterHub.
We recognize that Kubernetes has many deployment options. As a project team with limited resources to provide end user support, we rely on community members to share their collective Kubernetes knowledge and JupyterHub experiences.
We hope that you will use this section to share deployments with on a variety of infrastructure and for different use cases. There is also a community maintained list of users of this Guide and the JupyterHub Helm Chart.
Please submit an issue/pull request to add to this section. Thanks.
Tips and Snippets¶
This is a page to collect a few particularly useful patterns and snippets that help you interact with your Kubernetes cluster and JupyterHub. If there’s something that you think is generic enough (and not obvious enough) to be added to this page, please feel free to make a PR!
Kubernetes has a helper script that allows you to auto-complete commands
and references to objects when using
kubectl. This lets you
TAB-complete and saves a lot of time.
Follow the kubectl installation instructions for your platform to find the shell autocompletion instructions.
Helm also has an auto-completion script that lets you TAB-complete your commands when using Helm.
Oftentimes people manage multiple Kubernetes deployments at the same time.
kubectl handles this with the idea of “contexts”, which specify which
Kubernetes deployment you are referring to when you type
kubectl get XXX.
To see a list of contexts currently available to you, use the following command:
kubectl config get-contexts
This will list all of your Kubernetes contexts. You can select a particular context by entering:
kubectl config use-context <CONTEXT-NAME>
Specifying a default namespace for a context¶
If you grow tired of typing
namespace=XXX each time you type a kubernetes
command, here’s a snippet that will allow you set a default namespace for
a given Kubernetes context:
kubectl config set-context $(kubectl config current-context) \ --namespace=<YOUR-NAMESPACE>
The above command will only apply to the currently active context, and will
allow you to skip the
--namespace= part of your commands for this context.
Using labels and selectors with
Sometimes it’s useful to select an entire class of Kubernetes objects rather
than referring to them by their name. You can attach an arbitrary set of
labels to a Kubernetes object, and can then refer to those labels when
To search based on a label value, use the
arguments. For example, JupyterHub creates a specific subset of labels for all
user pods. You can search for all user pods with the following label query:
kubectl --namespace=<YOUR-NAMESPACE> get pod \ -l "component=singleuser-server"
For more information, see the Kubernetes labels and selectors page.
Asking for a more verbose or structured output¶
Sometimes the information that’s in the default output for
kubectl get <XXX>
is not enough for your needs, or isn’t structured the way you’d like. We
recommend looking into the different Kubernetes output options, which can be
modified like so:
kubectl --namespace=<NAMESPACE> get pod -o <json|yaml|wide|name...>
You can find more information on what kinds of output you can generate at the kubectl information page. (click and search for the text “Output Options”)
This is a community maintained list of organizations / people using the Zero to JupyterHub guide and Helm chart to maintain their JupyterHub. Send us a Pull Request to add yourself to this alphabetically sorted list!
Data Science Education Program’s DataHub at University of California, Berkeley
A group of universities near Paris (led by a team at CNRS and Polytechnique) deployed a cross-institutional JupyterHub for university use.
Globus runs an [instance of Zero-to-JupyterHub](<https://jupyter.demo.globus.org/) to help users learn about Globus REST APIs.