docker-jupyter

Dockerized Jupyter notebooks (Python, R, Spark) for data science explorations.

Introduction

Several Dockerfiles are included:

• base-jupyter: Contains the base Debian (stretch) system and everything to run a Jupyter Notebook server.
• r-jupyter: Extends base-jupyter with the R kernel, several R packages, and the ISLR library.
• python-jupyter: Extends base-jupyter with Python (2 and 3) environments focused on data science, loading common packages such as scikit-learn, pandas, seaborn, etc.
• pyspark-jupyter: Extends python-jupyter with Apache Spark.

Installation

To build any of the images, run the build script in that image's folder. The base-jupyter image must first be built and available on your system to build any other image. The pyspark-jupyter image depends on the python-jupyter image.

Running

Use the run.sh script in the corresponding image's folder. Then run the start-notebook.sh script, which is in the container root user's path. This provides a URL hash required to access the notebook, such as:

    Copy/paste this URL into your browser when you connect for the first time,
    to login with a token:
        http://localhost:8888/?token=3e75593fe1a03cf6540c1c4338b33a29e76399bc9fee8c4e

If running remotely, replace localhost with the appropriate IP address.

Options

The run.sh can mount two external directories, using the following options:

• -w|--workdir host_directory_path: Jupyter's working directory. This becomes /opt/workdir in the container.
• -d|--datadir host_directory_path: A directory where data is stored. This becomes /opt/datadir in the container. These options are useful for organization's sake, and also in the case when a working or data directory isn't located under your $HOME directory.

Known Issues

• If a browser is already running and attempting to contact the host on port 8888, the notebook server may ask for a password. If this happens, restart the container while ensuring that any existing browser tabs pointing at the server are closed.