JupyterHub

JupyterHub

Test Current Setup

Start service (but should already be running) and follow output

sudo systemctl start jupyterhub
sudo systemctl status jupyterhub
sudo journalctl -f -u jupyterhub

Open page with valid certificate https://banzhaf.chickenkiller.com:7000

Setup New Installation

• create new user and group jupyterhub and put user in group shadow
• allow sudo for members of group. Create /etc/sudoers.d/jupyterhub

Cmnd_Alias JUPYTER_CMD = /home/jupyterhub/miniconda3/bin/sudospawner
jupyterhub ALL=(%jupyterhub) NOPASSWD:JUPYTER_CMD

• install conda for that user

script=Miniconda3-latest-Linux-x86_64.sh
wget "https://repo.anaconda.com/miniconda/$script"
chmod +x "$script"
"./$script" -b
~/miniconda3/bin/conda init bash
. ~/.bashrc
conda update conda

• install packages (Minimum: jupyter hub, sudospawner and lab. Recommended: language server, c++ kernel and other useful stuff...)

conda install -c conda-forge jupyterhub jupyterlab jupyterlab-lsp python-lsp-server sudospawner xeus-cling opencv pandas matplotlib

• configure jupyterhub

jupyterhub --generate-config -f jupyterhub_config.py
add line "c.JupyterHub.bind_url = 'http://:7000'" to config file 

• start the hub

jupyterhub --JupyterHub.spawner_class=sudospawner.SudoSpawner

• login at http://localhost:7000

JupyterHub with SSL

super easy if you already have a cert and key (e.g. from letsencrypt): add these two lines to /home/jupyterhub/jupyterhub_config.py

c.JupyterHub.ssl_cert = '/etc/letsencrypt/live/banzhaf.chickenkiller.com/fullchain.pem'
c.JupyterHub.ssl_key = '/etc/letsencrypt/live/banzhaf.chickenkiller.com/privkey.pem'

JupyterHub as a Service

• create script to init conda environment and then start jupyterhub

#!/bin/bash
. ~/.bashrc
exec jupyterhub --JupyterHub.spawner_class=sudospawner.SudoSpawner

• create systemd script /etc/systemd/system/jupyterhub.service

Description=Jupyterhub

[Service]
User=jupyterhub
ExecStart=/home/jupyterhub/jupyterhub.sh
WorkingDirectory=/home/jupyterhub

[Install]
WantedBy=multi-user.target

• Install own kernel with conda environment
  • start terminal and create conda environment as usual (or list existing envs to choose from)
  • activate the conda environment and make sure ipykernel module is installed
  • register new kernel

myenv=my-new-env
conda activate "$myenv"
python -m ipykernel install --user --name "$myenv" --display-name "Python ($myenv)"

• • reload hub page and start Launcher (+) to see the new kernel