To have your own Jupyter Notebook Server running 24/7 on Raspberry Pi. If connected to your router, with port forwarding and DDNS set up, you can carry out your Jupyter tasks on the move.

This is a minimal Jupyter notebook server with Python 3.6.6 installed with Berryconda3 and resin/rpi-raspbian:latest as base image. JupyterLab is also installed. To use Lab interface just replace tree with lab in notebook URL.

• Minimal function Jupyter notebook 5.2.x installation
• Conda based Python 3.6.6 for Raspberry pi
• Unprivileged user jovyan in group users (gid=100) with ownership over /home/jovyan and /opt/conda
• Tini 0.18.0 as the container entrypoint

These packages are installed:

locales ca-certificates wget bzip2 vim fonts_liberation

Of course, first you will need to install Docker on your Raspberry Pi:

curl -sSL https://get.docker.com | sh

Then pull the container:

docker pull movalex/rpi-jupyter-conda:latest

Run in detached mode:

docker run -d -p 8888:8888 movalex/rpi-jupyter-conda

Now you can access your notebook at http://<docker host IP address>:8888

The image already has following configuration, so 'open browser' option is disabled and your notebook is accessible from any ip:

• c.NotebookApp.open_browser = False
• c.NotebookApp.ip = '*'

If you would like to change configuration, create your own jupyter_notebook_config.py (or use sample file from this repository), place it to $HOME/.jupyter folder and link it to docker volume. Run the following to map a local config file to the container:

docker run -it -p 8888:8888 -v <path to your config file>:/home/jovyan/.jupyter/jupyter_notebook_config.py movalex/rpi-jupyter-conda

To add a password for your notebook, you can alter c.NotebookApp.password = '' section in config file. You can add password later after jupyter is installed. Just run

from IPython.lib import passwd
passwd()

in Python Jupyter notebook to generate new password, and add output as a value of c.NotebookApp.password.

The notebook will run under unpriviledged user jovyan (uid=1000) with ownership of /home/jovyan and opt/conda. If you want to mount your default working directory on the host to preserve work even when notebook is not running or destroyed, use additional -v option:

docker run -it docker run -it -p <host port>:<dest port> -v <path to your config file>:/home/jovyan/.jupyter/jupyter_notebook_config.py -v /some/host/folder/for/work:/home/jovyan/work  movalex/rpi-jupyter-conda

To login a bash session use:

docker exec -it <container id> /bin/bash

If you want to start bash session with root accesss so you could do more, use this command:

docker exec -it -u 0 <container id> /bin/bash

This Docker container has only Python 3 kernel. If you still need Python2, you can add Python2 Conda environment and add it to Jupyter with ipykernel module. But since this module requires gcc and libzmq3, you'll need to install them first in priviledged bash session.

IMHO better solution would be using this datascience container with all necessary libraries (and many others) pre-installed. Here you can read how to add Python2 kernel to Jupyter Notebook.