jhub-client

JupyterHub Client

Automation of JupyterHub

Install

pip install jhub-client

Command Line Usage

Below are some example use cases of the tool. Note that with an admin api token you can impersonate users and create temporary users. Service api tokens do not have an associated user therefore must run as existing users or temporary users.

Set the api token used for jhub-client.

export JUPYTERHUB_API_TOKEN=<api-token>

Optionally basic authentication can be used instead of token based (not recommended but needed in some use cases). In which case username and password are required.

export JUPYTERHUB_USERNAME=<username>
export JUPYTERHUB_PASSWORD=<password>

Token

$ jhubctl token --help
usage: __main__.py token [-h] [--hub HUB] [--name NAME]

optional arguments:
  -h, --help   show this help message and exit
  --hub HUB    url for running jupyterhub cluster
  --name NAME  name to give to jupyterhub api token

A simple command line utility can be used to create a token. This only supports basic authentication (username and password) login to the jupyterhub cluster.

Run

$ jhubctl run --help
usage: __main__.py run [-h] -n NOTEBOOK [--auth-type {token,basic}] [--hub HUB] [-u USERNAME]
                       [--user-options USER_OPTIONS] [--temporary-user] [-d] [--stop-server] [--validate]
                       [--kernel-spec KERNEL_SPEC] [--output-filename OUTPUT_FILENAME]

optional arguments:
  -h, --help            show this help message and exit
  -n NOTEBOOK, --notebook NOTEBOOK
                        notebook to run
  --auth-type {token,basic}
                        jupyterhub authentication type to use with default of token based
  --hub HUB             url for running jupyterhub cluster with default of 'http://localhost:8000'
  -u USERNAME, --username USERNAME
                        username to run notebook as
  --user-options USER_OPTIONS
                        json object representing user server options
  --temporary-user      create user temporarily if does not exist
  -d, --daemonize       run notebook asyncronously
  --stop-server         stop server after completion of notebook
  --validate            validate notebook output matches
  --kernel-spec KERNEL_SPEC
                        kernel spec to launch is not specified will use default
  --output-filename OUTPUT_FILENAME
                        output filename for results of running notebook

Run notebook as given token user syncronously

You can run a given notebook as a pre-existing user syncronously. The api token either has to be for the given user or an admin token.

jhubctl --verbose run --notebook <notebook> --hub <hub_url>

Run notebook as given token user syncronously and validate notebook output matches

You can run a given notebook as a pre-existing user syncronously. The api token either has to be for the given user or an admin token.

jhubctl run --notebook <notebook> --hub <hub_url> --validate

Run notebook as given token user asyncronously and shutdown server after completion

You can run a given notebook as a pre-existing user asyncronously and stop server after completion. The api token either has to be for the given user or an admin token.

jhubctl run --notebook <notebook> --hub <hub_url> --daemonize --stop-server

Run notebook as given token user with user options

While this is an advanced use case, it is often times encountered with kubernetes jupyerhub clusters where you may want to select a given profile e.g. small, medium, gpu jupyterlab session. For these clusters you must supply --user-options='{"profile": 0}' where 0 is replaced with the index of the profile you would like to choose. It is possible for other more customized jupyterhub clusters that different options must be used.

jhubctl run --notebook <notebook> --hub <hub_url> --user-options='{"profile": 1}'

Run notebook as given token user with user options

You can run a given notebook with any available kernel spec.

jhubctl run --notebook <notebook> --hub <hub_url> --kernel-spec=python3

Run a given notebook as a temporary user

Additionally you can temporarily create a user user-<uuid> or supply the temporary user's username. The username will be deleted upon completion. The api token requires admin permissions.

jhubctl run --temporary-user --notebook <notebook> --hub <hub_url> [--username <username>]

Testing

Bring up test jupyterhub cluster

cd tests/assets
docker-compose up --build

Run unit tests

pytest

FAQ

Creating an API Token

Login to the given jupyterhub cluster

Access the hub control page. The url will be <hub_url>/hub/home.

Click on Token in top left corner and request new api token. This token will have the permissions of the user. Make sure to set the environment variable JUPYTERHUB_API_TOKEN to the value of the given token.

If you want to add a service token instead edit the jupyterhub configuration.

c.JupyterHub.services = [
 {
        'name': '<my-service-name>',
        'api_token': '<my-super-secret-long-token>',
        'oauth_no_confirm': True,
        'admin': True
    }
]

Code of Conduct

To guarantee a welcoming and friendly community, we require contributors to follow our Code of Conduct.

License

jhub-client is BSD-3 Licensed