Skip to content

Latest commit

 

History

History
76 lines (58 loc) · 3.49 KB

README.md

File metadata and controls

76 lines (58 loc) · 3.49 KB

wrapspawner for Jupyterhub

This package includes WrapSpawner and ProfilesSpawner, which provide mechanisms for runtime configuration of spawners. The inspiration for their development was to allow users to select from a range of pre-defined batch job profiles, but their operation is completely generic.

Installation

  1. Most users can install via pip:

    pip install wrapspawner

    To install an editable copy for development, from root directory of this repo (where setup.py is), run pip install -e . See also pip VCS support if you need a specific revision.

  2. Add lines in jupyterhub_config.py for the spawner you intend to use, e.g.

       c.JupyterHub.spawner_class = 'wrapspawner.ProfilesSpawner'
  3. Depending on the spawner, additional configuration will likely be needed.

Wrapper and Profile Spawners

Overview

WrapSpawner provides a mechanism to wrap the interface of a JupyterHub Spawner such that the Spawner class to use for single-user servers can be chosen dynamically. Subclasses may modify the class or properties of the child Spawner at any point before start() is called (e.g. from Authenticator pre_spawn hooks or options form processing) and that state will be preserved on restart. The start/stop/poll methods are not real coroutines, but simply pass through the Futures returned by the wrapped Spawner class.

ProfilesSpawner leverages JupyterHub's Spawner "options form" feature to allow user-driven configuration of Spawner classes while permitting:

  • configuration of Spawner classes that don't natively implement options_form
  • administrator control of allowed configuration changes
  • runtime choice of which Spawner backend to launch

Example

Here is a screenshot of a typical dropdown menu letting the user choose between several SLURM instances:

The following configuration snippet lets the user choose between a Jupyter server running as a local process or one of two different Docker Images to run within DockerSpawner.

c.JupyterHub.spawner_class = 'wrapspawner.ProfilesSpawner'
c.Spawner.http_timeout = 120
#------------------------------------------------------------------------------
# ProfilesSpawner configuration
#------------------------------------------------------------------------------
# List of profiles to offer for selection. Signature is:
#   List(Tuple( Unicode, Unicode, Type(Spawner), Dict ))
# corresponding to profile display name, unique key, Spawner class,
# dictionary of spawner config options.
# 
# The first three values will be exposed in the input_template as {display},
# {key}, and {type}
#
 c.ProfilesSpawner.profiles = [
       ( "Host process", 'local', 'jupyterhub.spawner.LocalProcessSpawner', {'ip':'0.0.0.0'} ),
       ('Docker Python 3', 'singleuser', 'dockerspawner.SystemUserSpawner', dict(image="jupyterhub/singleuser")),
       ('Docker Python 3 Scipy', 'scipy-notebook', 'dockerspawner.SystemUserSpawner', dict(image="jupyter/scipy-notebook")),
 ]

History

These mechanisms originated as part of the batchspawner package. The batchspawner README contains additional examples on the use of ProfilesSpawner.