An Introduction Docker for Research Software Engineers

SciPy 2015

Date

2015-07-07

Author

Matt McCormick <matt.mccormick@kitware.com>

Introduction

Docker is a new technology that has caught the software industry by storm. Docker is a new type of tool to easily build, ship, and run reproducible, binary applications. In this talk, we will introduce Docker from the perspective of a scientific research software engineer.

We will

1. Generate an understanding of what Docker is by comparing it to existing technologies.
2. Give an introduction to basic Docker concepts.
3. Describe how Docker fits into the scientific Python analysis workflow.

Run the Presentation Live, Locally -------------------------------

All the software required to run the presentation is available in the provided Docker image. This image is based of the image presented on Modern Scientific Computing with Python at IEEE NSS/MIC 2014.

Install Docker

To install Docker, follow the online installation instructions.

Import the Docker Image

There are two options to import the Docker image.

Pull the image from DockerHub.

DockerHub is an online repository of docker images that makes it easy to search, push, and pull images from public or private repositories. To download the image from DockerHub:

docker pull thewtex/scipy-2015-docker-presentation

If there are permission denied errors, make sure your user is in the docker group:

sudo gpasswd --add $(whoami) docker
sudo service docker restart
newgrp

or see more instructions for Ubuntu.

Build the image from its sources.

Docker images are created from a set of instructions that are stored in a Dockerfile. These are the set of commands that are used to set up the image, like installing packages, compiling dependencies, editing configurations, etc. To build the image:

git clone https://github.com/thewtex/scipy-2015-docker-presentation
cd scipy-2015-docker-presentation
./build.sh

Run the Docker Image

To run the IPython notebook server, clone the Git repository if you have not already done so:

git clone https://github.com/thewtex/scipy-2015-docker-presentation
cd scipy-2015-docker-presentation

Then run the run.sh script with a password of your choice:

./run.sh MakeAPassword

This command should be executed from the directory containing the *.ipynb notebook files so $PWD will make the notebook files available to the docker container. Replace MakeAPassword with a password of your choice.

Docker container port 8888, where the IPython notebook server is listening for connections, will be forwarded to the localhost on port 443. To connect to the notebook server, point your browser to https://localhost/.

Note

On OSX and Windows, extra steps are required to find the address and the port that the container has been forwarded to. In a Boot2Docker shell, run:

boot2docker ip

This should return an address like 192.168.59.103. Next, find the port in the docker environment with:

docker port notebook 8888

This should return a port like 49153.

Port your browser to the resulting address and port; for example: https://192.168.59.103:49153.

After connecting to the container, an "untrusted connection" warning from your browser is expected.

To list the running containers:

docker ps

To list the running and stopped containers:

docker ps -a

To stop the container:

docker stop techlunch

To start the container again:

docker start techlunch

To remove the container:

docker rm techlunch