Zero to JupyterHub with Kubernetes

General
1

Manage to setup this deployment on ubuntu microk8s easily with helm couple of queries on how to further customize the setup

  1. in the config.yaml file, i am using Active directory as the default security authentication. How do i parse the account credential as a secret and not in clear text?

search:
filter: ‘({login_attr}={login})’
user: ‘ldap_search_user_technical_account’
password: ‘secret’

  1. In order to run gpu on the notebook, i would need to use a docker image that has nvidia docker inside? i saw a docker image from cschranz/gpu-jupyter and i will be using that… on my microk8s level do i still have to enable gpu addons?

  2. Let say i have a user who is using a minimal notebook image and realize he requires more frameworks to be installed, we managed to sync down a internal pypi mirror, how can we tweak to pip install from that mirror? we have to recompile our docker image to include to point the pip repo in /etc/pip.conf?

2

For qn2 how do i limit the shares to the gpu card too? 1 gpu card per notebook? If i have 1 gpu card only how is this being shared if multiple gpu notebook pods are spun up?