Note: JupyterHub with JupyterLab Install using Conda

Table of Contents

Introduction

This is a quick note about setting up a JupyterHub server and JupyterLab using conda with Anaconda Python.

A few weeks ago I posted Note: How To Install JupyterHub on a Local Server In that post I used the system Python3 and a virtenv with pip to install JupyterHub. That is a perfectly fine way to do the install but I think I prefer what is presented in this new post using conda.

If you want to do this install yourself I recommend that you read through the post linked above since it contains more detail and discussion.

This JupyterLab setup could be useful for sharing a server or workstation with a few other users. It could be used as a remote resource from a browser running on a laptop or workstation. It is not configured for public network access but would be useful over a VPN connection to "the office".

Install conda (globally)

Install extra packages

I needed to add curl on my Ubuntu 20.04 server test system, you may need other packages.

sudo apt-get install curl

Add apt repo for conda

curl -L https://repo.anaconda.com/pkgs/misc/gpgkeys/anaconda.asc | sudo apt-key add -

sudo echo "deb [arch=amd64] https://repo.anaconda.com/pkgs/misc/debrepo/conda stable main" | sudo tee  /etc/apt/sources.list.d/conda.list

Install

This installs conda (miniconda Python environment) to /opt/conda

sudo apt-get update
sudo apt-get install conda

Setup PATH and Environment for conda when users login

sudo ln -s /opt/conda/etc/profile.d/conda.sh /etc/profile.d/conda.sh

Install JupyterHub with conda

Sudo to a root shell for the install and setup

sudo -s

source the conda env for root

Using the "dot" command,

. /etc/profile.d/conda.sh

Create a conda "env" for JupyterHub with JupyterLab and install

We will want ipywidgets installed here too.

conda create --name jupyterhub   jupyterhub jupyterlab ipywidgets

Create JupyterHub config file

Note: you will need to use full path to jupyterhub executable in the env.

mkdir -p /opt/conda/envs/jupyterhub/etc/jupyterhub
cd /opt/conda/envs/jupyterhub/etc/jupyterhub
/opt/conda/envs/jupyterhub/bin/jupyterhub --generate-config

Set default config to use JupyterLab

sed -i "s/#c.Spawner.default_url = ''/c.Spawner.default_url = '/lab'/" jupyterhub_config.py

Use systemd to start jupyterhub on boot

Create a systemd "Unit" file for starting JupyterHub,

sudo mkdir -p /opt/conda/envs/jupyterhub/etc/systemd

sudo cat << EOF > /opt/conda/envs/jupyterhub/etc/systemd/jupyterhub.service

[Unit]
Description=JupyterHub
After=syslog.target network.target

[Service]
User=root
Environment="PATH=/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/opt/conda/envs/jupyterhub/bin"
ExecStart=/opt/conda/envs/jupyterhub/bin/jupyterhub -f /opt/conda/envs/jupyterhub/etc/jupyterhub/jupyterhub_config.py

[Install]
WantedBy=multi-user.target

EOF

Link to OS systemd directory

sudo ln -s /opt/conda/envs/jupyterhub/etc/systemd/jupyterhub.service /etc/systemd/system/jupyterhub.service

Start JupyterHub, and enable it as a service,

systemctl start jupyterhub.service 
systemctl enable jupyterhub.service

That's it! You now have JupyterHub installed and running. It will be running on the ip address for the system you installed onto using port 8000. Any user with a login account on the server will be able to access JupyterLab and use the machine.

There will be a default Python3 env available but will likely want to add some other "env's" for users. You can add env's for all users as in the examples below.

System-wide kernels are located in,

/usr/local/share/jupyter/kernels/

Individual users can create their own custom env's using conda in their home directory. Users env's will be located in

$HOME/.local/share/jupyter/kernels/

Add some extra (system-wide) kernels for JupyterLab

Anaconda3 (full package)

sudo /opt/conda/bin/conda create --yes --name anaconda3 anaconda ipykernel
sudo /opt/conda/envs/anaconda3/bin/python -m ipykernel install --name 'anaconda3' --display-name "Anaconda3"

TensorFlow 2.1 GPU

If your server has NVIDIA CUDA capable GPU's and the NVIDIA driver installed then you can add GPU accelerated env's.

sudo /opt/conda/bin/conda create --yes --name tensorflow2.1-gpu tensorflow-gpu ipykernel
sudo /opt/conda/envs/tensorflow2.1-gpu/bin/python -m ipykernel install --name 'tensorflow2.1-gpu' --display-name "ThensoFlow 2.1 GPU"

PyTorch 1.4 GPU

sudo /opt/conda/bin/conda create --yes --name pytorch1.4-gpu ipykernel pytorch torchvision  cudatoolkit=10.1  -c pytorch 
sudo /opt/conda/envs/pytorch1.4-cpu/bin/python -m ipykernel install --name 'pytorch1.4' --display-name "PyTorch 1.4 CPU"

Enjoy!

Happy computing! and please stay safe, best wishes –Don @dbkinghorn