Linux

The following script will install Jupyter Hub in single user mode (i.e. only a single Linux user can login to Jupyter Hub using the web interface).

Prerequisites

First install Python & PIP, then NodeJS, then Jupyter Hub, then configurable-http-proxy :

curl -sL https://deb.nodesource.com/setup_14.x | sudo -E bash -
sudo apt-get install -y nodejs python3-pip
sudo pip3 install jupyterhub
sudo npm i -g configurable-http-proxy

Installing the Jupyter Hub systemd service

Run the following script using sudo!

#!/bin/bash
# This script installs and enables/starts the JupyterHub systemd service
export NAME=JupyterHub

# Create service file
cat >/etc/systemd/system/${NAME}.service <<EOF
[Unit]
Description=${NAME}

[Service]
Type=simple
ExecStart=/usr/bin/env jupyterhub --port=11569
User=root
Group=root

Restart=on-failure
RestartSec=5s

[Install]
WantedBy=multi-user.target
EOF

# Enable and start service
systemctl enable --now ${NAME}

This script will install a systemd service named JupyterHub autostart it on boot. It is running on port 11569 by default.

Note: This will only allow a single (preconfigured) user to login to Jupyter lab! See How to run Jupyter Hub (multi-user mode) using systemd on how to deploy Jupyter Hub in multi-user mode using systemd, which allows any Unix user to login!

The following script will install Jupyter Hub in single user mode (i.e. only a single Linux user can login to Jupyter Hub using the web interface).

Prerequisites

First install Python & PIP, then NodeJS, then Jupyter Hub, then configurable-http-proxy :

curl -sL https://deb.nodesource.com/setup_14.x | sudo -E bash -
sudo apt-get install -y nodejs python3-pip
sudo pip3 install jupyterhub
sudo npm i -g configurable-http-proxy

Installing the Jupyter Hub systemd service

Run the following script as the user you want to be able to login! Do not run the script using sudo !

#!/bin/bash
# This script installs and enables/starts a systemd service
export NAME=JupyterHub-$USER
export GROUP=$(id -gn $USER)
# Create service file
sudo tee /etc/systemd/system/${NAME}.service <<EOF
[Unit]
Description=${NAME}

[Service]
Type=simple
ExecStart=/usr/bin/env jupyterhub

WorkingDirectory=$HOME
User=$USER
Group=$GROUP

Restart=on-failure
RestartSec=5s

[Install]
WantedBy=multi-user.target
EOF

# Enable and start service
systemctl enable --now ${NAME}

This script will install a systemd service named JupyterHub-$USER (where $USER is the current user, e.g. uli) and autostart it on boot.

Running multiple services

If you run multiple services, you can run the script for each user and choose a unique port for each service by adding --port=7219 to the /usr/bin/env jupyter hub command, e.g.

ExecStart=/usr/bin/env jupyterhub --port=7192

Alternatively, you can run a single systemwide Jupyter Hub in multi-user mode where multiple users can log in.

If you want to run your Jupyter Lab as a network service on any modern Linux distribution, you can install a systemd service that runs Jupyter. First, you need to install jupyter lab using

sudo pip3 install jupyterlab

In case you don’t have pip3, use sudo apt -y install python3-pip or the equivalent on your distribution.

Note that this script will run Jupyter without token authentication and without password and it will listen on any IP (--ip=0.0.0.0) by default. Either change the command line flags or be aware of the security implications !

#!/bin/bash
# This script installs and enables/starts a systemd service
export NAME=Jupyter

# Create service file
cat >/etc/systemd/system/${NAME}.service <<EOF
[Unit]
Description=${NAME}

[Service]
Type=simple
ExecStart=/usr/bin/env jupyter lab --ip=0.0.0.0 --port 17256 --LabApp.token=''

WorkingDirectory=/home/uli/jupyter
User=uli
Group=uli

Restart=on-failure
RestartSec=5s

[Install]
WantedBy=multi-user.target
EOF

# Enable and start service
systemctl enable --now ${NAME}

You need to change the following entries in the script in order to make it work for you:

WorkingDirectory=/home/uli/jupyter
User=uli
Group=uli

Set the WorkingDirectory to whatever directory you want Jupyter to run in. Note that anyone being able to access the webinterface will basically have full access to that directory!

Set User and Group to the user that should run. Note that running Jupyter as root is not allowed. In case you still want to do it, add the --allow-root flag to the command line options.

Now run the script as root to install the service:

sudo ./install-jupyter-service.sh

Now you can access Jupyter at http://[ip of the computer]:17256.

Changing the configuration

In order to change the configuration, I recommend to edit /etc/systemd/systemd/Jupyter.service (or /etc/systemd/systemd/${NAME}.service if you changed export NAME=Jupyter) directly. After that, run

sudo systemctl daemon-reload
sudo systemctl restart Jupyter

You can also change the installation script and re-run it, but you still need to run daemon-reload and restart.

Running multiple Jupyter instances

In order to run  multiple instances, just run multiple copies of the installation script with different names. For example, use

export NAME=Jupyter-DeepLearning

Debugging Jupyter Lab output

If you have problems with Juypter Lab starting up, use

sudo systemctl status Jupyter

in order to view the status and

sudo journalctl -xfu Jupyter

to view all the logs.

Status output example:

● Jupyter.service - Jupyter
     Loaded: loaded (/etc/systemd/system/Jupyter.service; enabled; vendor preset: enabled)
     Active: active (running) since Fri 2021-06-11 03:44:28 CEST; 4s ago
   Main PID: 48753 (jupyter-lab)
      Tasks: 1 (limit: 14226)
     Memory: 51.7M
     CGroup: /system.slice/Jupyter.service
             └─48753 /usr/bin/python3 /usr/local/bin/jupyter-lab --ip=0.0.0.0 --port 17256 --LabApp.token=

Jun 11 03:44:29 uli-desktop env[48753]: [I 2021-06-11 03:44:29.215 ServerApp] nbclassic | extension was successfully loaded.
Jun 11 03:44:29 uli-desktop env[48753]: [I 2021-06-11 03:44:29.216 LabApp] JupyterLab extension loaded from /usr/local/lib/python3.8/dist-packages/jupyterlab
Jun 11 03:44:29 uli-desktop env[48753]: [I 2021-06-11 03:44:29.216 LabApp] JupyterLab application directory is /usr/local/share/jupyter/lab
Jun 11 03:44:29 uli-desktop env[48753]: [I 2021-06-11 03:44:29.219 ServerApp] jupyterlab | extension was successfully loaded.
Jun 11 03:44:29 uli-desktop env[48753]: [I 2021-06-11 03:44:29.220 ServerApp] Serving notebooks from local directory: /dev/shm
Jun 11 03:44:29 uli-desktop env[48753]: [I 2021-06-11 03:44:29.220 ServerApp] Jupyter Server 1.8.0 is running at:
Jun 11 03:44:29 uli-desktop env[48753]: [I 2021-06-11 03:44:29.220 ServerApp] http://uli-desktop:17256/lab
Jun 11 03:44:29 uli-desktop env[48753]: [I 2021-06-11 03:44:29.220 ServerApp]     http://127.0.0.1:17256/lab
Jun 11 03:44:29 uli-desktop env[48753]: [I 2021-06-11 03:44:29.220 ServerApp] Use Control-C to stop this server and shut down all kernels (twice to skip confirmation).
Jun 11 03:44:29 uli-desktop env[48753]: [W 2021-06-11 03:44:29.224 ServerApp] No web browser found: could not locate runnable browser.

Uninstalling the Jupyter Lab service

In order to just stop and disable autostart (but not uninstall) the jupyter lab service, use

sudo systemctl disable --now Jupyter

After that, you can just remove the service file in order to permanently uninstall the service:

sudo rm /etc/systemd/system/Jupyter.service

You can always reinstall using our installation script.

Note that if you have changed the export NAME=... line, you need to replace Jupyter by the value of Name