mdocker

An easy-to-use wrapper for multi-platform Docker image builds.

Contents

• mdocker
  • Contents
  • Description
  • Usage
  • Package Compatibility
  • Examples
  • Installation
    • From PyPI (recommended)
    • Local from source
    • No installation, direct run from source
  • License

Description

mdocker is a simple wrapper over Docker Buildx, which can be used for convenient image builds targeted for multiple platforms.

This tool was originally designed as a workaround to a limitation that Buildx has with the --load parameter.

The amount of target platforms specified for this wrapper is equal to the amount of tags generated in local cache.

E.g., if linux/arm64 and linux/amd64 were specified as target platforms for the demo image, you will get demo:arm64 and demo:amd64 built and stored within your local Docker cache.

Usage

mdocker requires an installation of Python 3.10+.

Below is a help message with the description of arguments.

$ python3 -m mdocker --help
usage: [-h] [--context BCONTEXT] [--file DFILE] [--platforms PLATFORMS] [--push] name

positional arguments:
  name                  specify a name for the image

options:
  -h, --help            show this help message and exit
  --context BCONTEXT    specify a path to build context
  --file DFILE          specify a path to Dockerfile
  --platforms PLATFORMS
                        specify target platforms (e.g., --platforms linux/amd64,linux/arm64)
  --push                push image to remote registry

Package Compatibility

In some cases, packages for Docker images may not be available across all target platforms. E.g., a package "some_package" may be available for amd64, but may not be available for arm64 platform.

Is cases like this, you can define custom installation rules within your Dockerfile itself.

An example of such custom rule usage can be found here.

Docker's default back-end (BuildKit) provides built-in variables that can be used for determining target platform set for the current build.

The full list of these variables can be found here.

Examples

You can try out mdocker with it's own Dockerfile!

The command below will build the demo image for both amd64 and arm64 target architectures:

python3 -m mdocker demo --platforms linux/amd64,linux/arm64

Optionally, you can specify paths to the build context and Dockerfile:

python3 -m mdocker demo --context . --file ./Dockerfile --platforms linux/amd64,linux/arm64

Installation

From PyPI (recommended)

To install latest mdocker package from PyPI, use:

python3 -m pip install mdocker

Local from source

To install and debug mdocker locally, in the root of repository use:

python3 -m pip install -e .

No installation, direct run from source

To run mdocker without any installation into local cache, in the root of repository use:

export PYTHONPATH=$(pwd)
python3 -m poetry install --no-root
python3 mdocker <arguments>

License

MIT