Containers¶

With many different operating systems and local setups, it is often a challenging task to get all the required software up and running. Containers are one possible solution that. They encapsulate the entire working environment including not only needed applications and packages, but also an operating system. Their main goal is to handle all dependencies in a standardized and reliable manner and ensure that the same applications can be run on the new computing environment (laptop, server, HPC, etc. depending on the application).

Docker¶

Docker is one container platform we have used in projects at the Centre.

Docker defines several types of objects, but here we will mention only images and containers. Docker containers are the same as containers discussed above: the full computing environments taking care of all dependencies. Docker images, on the other hand, are templates defining how these environments are created.

Install Docker¶

To install Docker, follow the official Docker installation guide.

Run a Docker container locally¶

Note

🌿

Aim: create and run a Docker container from a publicly available Docker image

Level: apprentice 🌿

Once Docker is installed and running locally, check the list of available containers by running the following command from the terminal:

docker ps -a

If you just installed Docker, this list will be empty. If there are any containers, the list will show container ID, image from which it was created, when it was created, current status, etc.

Copy the following command in the terminal to pull a public Docker image, create a container from it and run a sample analysis. As an example, we here use immuneML image. immuneML is a software for machine learning analysis of immune receptors. In this example run, it will generate a synthetic dataset and try to fit a classifier to predict if a person is sick or healthy for the simulated disease.

This command will:

make the current working directory (pwd) visible inside the container at path /data
name the container my_container
download the immuneML Docker image from DockerHub from user milenapavlovic
in a thus created environment, it will run immune-ml-quickstart as described above and store the output to /data/output/ directory which will be visible later in the current working directory under output/.

docker run -it -v $(pwd):/data --name my_container milenapavlovic/immuneml immune-ml-quickstart /data/output/

The image reference (here milenapavlovic/immuneml) can be replaced with any relevant image, and the command following the image should then be adapted to the particular use-case for the downloaded image.

Create a Docker image and publish it on DockerHub¶

Note

🌳

Aim: provide references for creating a Docker image and publishing it in a public repository DockerHub

Level: journeyman 🌳

To create a Docker image, follow the Getting started guide. Parts 1-4 describe how to come up with an image and publish it on DockerHub, that can then be used as described above.

If you are using GitHub for code development and want to use that code as a basis for a Docker image, this process can also be automated to create a new image on a certain event and publish it on DockerHub.

For example, see the action YAML file on immuneML repository. Every time push is performed to the master branch, the image will be rebuilt based on the Dockerfile at the repository and published.