Anaconda-Notebook

IPython 3 Notebook and Terminal + Anaconda

Screencast Demo

Overview

A docker image with the intent to provide an IPython 3/Jupyter notebook and terminal access, similar to what is provided with jupyter/demo, except with a full Anaconda install, owned by the user. The benefits are that you can use conda or pip to quickly install new packages via the terminal.

A python 3 install is used, but has a python 2 IPython kernel that is tied to a python 2 conda environment, named python2. The python 2 environment is bare, with only python, pip, IPython and associated requirements.

The goal is to provide a sandbox IPython environment with most of what you'd need, but still provide the capability to install things as needed.

See the demo, it shows running it on OSX via boot2docker, running the included sample notebook, then installing seaborn with Conda, then running a seaborn example.

Included Python Packages

See Anaconda Package Documentation

Modifications to Standard Anaconda Install

• 'python2' environment
• IPython Startup Script
  • Sets matplotlib nbagg backend (bypasses qt dependencies), then sets matplotlib output mode to inline. This makes sure we always output to notebook by default.

Getting Started

Install Docker

• (Linux) Docker
• (Mac/Windows) Boot2Docker

Downloading the Latest Image

docker pull rothnic/anaconda-notebook

Running the Docker Image

If using with tmpnb, see their readme for using this image. Some example commands are provided below, but they may fall out of sync with tmpnb if it is updated, so be sure to check the project.

# create a bunch of unique keys to identify the workers
export TOKEN=$( head -c 30 /dev/urandom | xxd -p )

# run the proxy, detached
docker run --net=host -d -e CONFIGPROXY_AUTH_TOKEN=$TOKEN --name=proxy jupyter/configurable-http-proxy --default-target http://127.0.0.1:9999

# run the orchestrator, detached, using anaconda-notebook
docker run --net=host -d -e CONFIGPROXY_AUTH_TOKEN=$TOKEN -v /var/run/docker.sock:/docker.sock --name=tmpnb jupyter/tmpnb python orchestrate.py --image='rothnic/anaconda-notebook' --command='/home/condauser/anaconda3/bin/ipython notebook --NotebookApp.base_url=/{base_path} --ip=0.0.0.0 --port {port}'

Otherwise, standalone use would be with:

docker run -p 8888:8888 -i -t rothnic/anaconda-notebook

Overview of run options:

• (required) -p 8888:8888 Publishes the container᾿s port or a range of ports to the docker host. If not used, the notebook container will not receive requests its port
• -i Keeps STDIN open
• -t Allocates a terminal, otherwise with just -i you get plain text output and can't ctrl-c to stop the run command
• -v <host directory>:<guest directory> Mounts a host directory into the container. Use this to utilize your own local directory of ipython notebooks

Note: This run command could be modified for your purposes, see docker run. For example, using -d will run the image detached, so that a terminal does not have to stay open.

Anaconda Environment

• The anaconda install is located at /home/condauser/anaconda3
• The notebook environment is located at /home/condauser/notebooks
• Access the notebook by navigating to <docker host ip>:8888

Installing New Packages

1. Open terminal from IPython tree view

   1a. (Python 2 Only) Type source activate python2

2. Type conda install <your package name>

   2a. If conda does not have your package, try: pip install <your package name>

Development

There are a few ways you could customize this image. You can either fork this project and make the changes directly, fork this project, create a new branch, and submit a pull request, or use the image as a base for your own. If you just want to add in your own notebooks, the easiest approach would be to use this as a base for your own image.

Extending

Create your own Dockerfile and add the following to the top of the file:

FROM rothnic/anaconda-notebook

Next, just copy in your notebooks into /home/condauser/notebooks, or apply whatever other changes you'd like. However, if you only need to include your own notebooks, and no other changes are required, then try this complete run command, instead of making modifications to the dockerfile:

docker run -i -t -p 8888:8888 -v \
<path to your ipython notebooks on host>:/home/condauser/notebooks rothnic/anaconda-notebook \
/home/condauser/anaconda3/bin/ipython notebook

Building

Either clone this repository, or extend it as mentioned above, then navigate to the directory where the Dockerfile exists, then enter the following command:

docker build .

Avoid Repeated Anaconda Download

The install.sh file will look for a file named Anaconda.sh that will be in the /src directory. You can download the Anaconda installer, rename it, and place it there so that it isn't downloaded each time you try to build.