Docker Resources

What is Docker?

Docker is a tool for packaging up operating systems, scripts, and environments in order to be able to run the same code regardless of what machine the code is executing on. This packaged code is called an image. There are three parts to Docker: a mechanism for building images, an image repository called Docker Hub, and a way to execute code in an image called a container. For using Batch effectively, we’re only going to focus on building images.


You can install Docker by following the instructions for either Macs or for Linux.

Creating a Dockerfile

A Dockerfile contains the instructions for creating an image and is typically called Dockerfile. The first directive at the top of each Dockerfile is FROM which states what image to create this image on top of. For example, we can build off of ubuntu:22.04 which contains a complete Ubuntu operating system, but does not have Python installed by default. You can use any image that already exists to base your image on. An image that has Python preinstalled is python:3.6-slim-stretch and one that has gcloud installed is google/cloud-sdk:slim. Be careful when choosing images from unknown sources!

In the example below, we create a Dockerfile that is based on ubuntu:22.04. In this file, we show an example of installing PLINK in the image with the RUN directive, which is an arbitrary bash command. First, we download a bunch of utilities that do not come with Ubuntu using apt-get. Next, we download and install PLINK from source. Finally, we can copy files from your local computer to the docker image using the COPY directive.

FROM 'ubuntu:22.04'

RUN apt-get update && apt-get install -y \
    python3 \
    python3-pip \
    tar \
    wget \
    unzip \
    && \
    rm -rf /var/lib/apt/lists/*

RUN mkdir plink && \
    (cd plink && \
     wget && \
     unzip && \
     rm -rf

# copy single script
COPY /scripts/

# copy entire directory recursively
COPY . /scripts/

For more information about Dockerfiles and directives that can be used see the following sources:

Building Images

To create a Docker image, use

docker build -t<my-project>/<my-image>:<tag> -f Dockerfile .

* `<dir>` is the context directory, `.` means the current working directory,
* `-t <name>` specifies the image name, and
* `-f <dockerfile>` specifies the Dockerfile file.
* A more complete description may be found `here: <>`__.

For example, we can build an image named<my-project>/<my-image> based on the Dockerfile named Dockerfile, using the current working directory as the context:

docker build -t<my-project>/<my-image>:<tag> -f Dockerfile .

In this example we prepend the image name with<my-project>/ so that it may be pushed to the Google Container Registry, in the next step.

Pushing Images

To use an image with Batch, you need to upload it somewhere Batch can read it: either the Google Container Registry or Docker Hub. Below is an example of pushing the image to the Google Container Registry. For more information about pulling and pushing images for the Google Container Registry, see here.

docker push<my-project>/<my-image>:<tag>

Now you can use your Docker image with Batch to run your code with the method BashJob.image() specifying the image as<my-project>/<my-image>:<tag>!