Jupyterhub: where is get_config defined and how do I create a custom Authenticator?

Question

I want to create a jupyterhub installation that uses a custom redirect to authenticate the user. The user would enter the url to our Jupyterhub, be redirected to a separate authentication system, then returned to the app, bypassing login.html. It seems initially a lot of hacking would be needed on the actual Jupyterhub source code to make this work.

From what I understand currently we would need customize jupyterhub/jupyterhub/handlers/login.py jupyterhub/auth.py. We wanted to avoid that and use the officially supported mechanisms. I thought maybe I could create my own Authenticator for use in a config file as in

c = get_config()
c.Authenticator.stuffGoesHere

and implement my own:

class LoginHandler(BaseHandler)  // login.py

in addition to

class Authenticator(LoggingConfigurable):  // auth.py

But I don't even see clearly how to do this as I cannot figure out where

 get_config()

is implemented and how to override this. Am I overthinking this? Whats the solution to wanting a separate system to handle the login (we can get the username from this external system for docker provisioning etc for dockerspawner). Btw, where is get_config actually defined?

Answer 1

The get_config is defined here: https://github.com/ipython/traitlets/blob/c5c9166373041dd561d8ad86e8990ae844d26306/traitlets/config/loader.py#L483-L501

And the reason you can use it directly without importing first is py3compat.execfile.

Back to your need, what you should do is create your CustomizedAuthenticator class which extends the jupyterhub.auth.Authenticator. And use it with:

c.JupyterHub.authenticator_class = CustomizedAuthenticator