dockerma 0.5.2

DockerMA facilitates building multi-arch containers with minimal fuss

DockerMA facilitates building multi-arch containers with minimal fuss.

What does it do?

dockerma aims to be a drop-in replacement for docker build and docker push to handle building what is often referred to as “multi-arch” images. A “multi-arch” image is actually a manifest, listing which actual image to use for a given architecture. The way to create these are still somewhat cumbersome, and considered experimental, especially if you are using a base image that is already “multi-arch”.

dockerma build works by reading your Dockerfile and looking for a specially formatted comment that selects which architectures you want to build. It will then create a manipulated version of your Dockerfile, to inject some tools for cross-building, and then build architecture specific images. If you have selected amd64 and arm, and build myapp:v1.0, the build will produce myapp:v1.0-amd64 and myapp:v1.0-arm.

dockerma push will take the architecture specific images that was built previously, and push them to the remote registry. It will then create a manifest, listing each of the images for the wanted architectures, and push the manifest under the requested tag. This will create a “multi-arch” image.

Supported architectures is based on what is available in the tooling used. The cross-building tools are from https://github.com/monsonnl/qemu-wrap-build-files, and at this writing only supports amd64, arm and arm64. If you want dockerma to support other architectures, you need to persuade qemu-wrap-build-files to support it first.

How do I use it?

Since dockerma aims to be a drop-in replacement, using it should be fairly simple.

The first step is to add the dockerma comment to your Dockerfile:

# dockerma archs:arm,amd64,arm64:

When you previously used docker build -t myapp:latest -t myapp:v1.0 . to build latest and v1.0 tags for your current architecture, you will now simply replace docker with dockerma:

dockerma build -t myapp:latest -t myapp:v1.0 .

Similary, where you used docker push -t myapp:latest, you now use dockerma:

dockerma push myapp:latest

What are the requirements?

• You need to enable experimental features in docker. This can be done by setting DOCKER_CLI_EXPERIMENTAL to enabled, or changing your docker configuration. Consult the docker documentation for details.
• Your build host must support running qemu-<arch>-static.
• Your base images needs to already support multi-arch (and in particular, the architectures you want to support). Many official images already do.

Are there any downsides to this approach?

• All images will have some additional files from the cross-build tooling, located in /cross-build.

Releasing a new version

1. Tag the new version: hg tag v1.2.3
2. Push the tag to Bitbucket
3. Watch the pipeline publish the new version to PyPI

New changes in v0.5.2

• 3ec7f836a1c1: Minor fixes
• ddf8faba9a65: Fix deployment step
• f2c2d66cb4e9: Configure pipeline to support push e2e test
• 7b7c6d7b7c05: Add e2e test for push