Skip to content

MATURE: A set of Maven tools for dealing with Dockerfiles

License

Notifications You must be signed in to change notification settings

ChinaJeckXu/dockerfile-maven

 
 

Repository files navigation

Dockerfile Maven

Build Status Maven Central License

Status: mature

At this point, we're not developing or accepting new features or even fixing non-critical bugs.

This Maven plugin integrates Maven with Docker.

The design goals are:

  • Don't do anything fancy. Dockerfiles are how you build Docker projects; that's what this plugin uses. They are mandatory.
  • Make the Docker build process integrate with the Maven build process. If you bind the default phases, when you type mvn package, you get a Docker image. When you type mvn deploy, your image gets pushed.
  • Make the goals remember what you are doing. You can type mvn dockerfile:build and later mvn dockerfile:tag and later mvn dockerfile:push without problems. This also eliminates the need for something like mvn dockerfile:build -DalsoPush; instead you can just say mvn dockerfile:build dockerfile:push.
  • Integrate with the Maven build reactor. You can depend on the Docker image of one project in another project, and Maven will build the projects in the correct order. This is useful when you want to run integration tests involving multiple services.

This project adheres to the Open Code of Conduct. By participating, you are expected to honor this code.

See the changelog for a list of releases

Set-up

This plugin requires Java 7 or later and Apache Maven 3 or later (dockerfile-maven-plugin <=1.4.6 needs Maven >= 3, and for other cases, Maven >= 3.5.2). To run the integration tests or to use the plugin in practice, a working Docker set-up is needed.

Example

For more examples, see the integration test directory.

In particular, the advanced test showcases a full service consisting of two micro-services that are integration tested using helios-testing.

This configures the actual plugin to build your image with mvn package and push it with mvn deploy. Of course you can also say mvn dockerfile:build explicitly.

<plugin>
  <groupId>com.spotify</groupId>
  <artifactId>dockerfile-maven-plugin</artifactId>
  <version>${dockerfile-maven-version}</version>
  <executions>
    <execution>
      <id>default</id>
      <goals>
        <goal>build</goal>
        <goal>push</goal>
      </goals>
    </execution>
  </executions>
  <configuration>
    <repository>spotify/foobar</repository>
    <tag>${project.version}</tag>
    <buildArgs>
      <JAR_FILE>${project.build.finalName}.jar</JAR_FILE>
    </buildArgs>
  </configuration>
</plugin>

A corresponding Dockerfile could look like:

FROM openjdk:8-jre
MAINTAINER David Flemström <dflemstr@spotify.com>

ENTRYPOINT ["/usr/bin/java", "-jar", "/usr/share/myservice/myservice.jar"]

# Add Maven dependencies (not shaded into the artifact; Docker-cached)
ADD target/lib           /usr/share/myservice/lib
# Add the service itself
ARG JAR_FILE
ADD target/${JAR_FILE} /usr/share/myservice/myservice.jar

Important note

The most Maven-ish way to reference the build artifact would probably be to use the project.build.directory variable for referencing the 'target'-directory. However, this results in an absolute path, which is not supported by the ADD command in the Dockerfile. Any such source must be inside the context of the Docker build and therefor must be referenced by a relative path. See spotify#101

Do not use ${project.build.directory} as a way to reference your build directory.

What does it give me?

There are many advantages to using this plugin for your builds.

Faster build times

This plugin lets you leverage Docker cache more consistently, vastly speeding up your builds by letting you cache Maven dependencies in your image. It also encourages avoiding the maven-shade-plugin, which also greatly speeds up builds.

Consistent build lifecycle

You no longer have to say something like:

mvn package
mvn dockerfile:build
mvn verify
mvn dockerfile:push
mvn deploy

Instead, it is simply enough to say:

mvn deploy

With the basic configuration, this will make sure that the image is built and pushed at the correct times.

Depend on Docker images of other services

You can depend on the Docker information of another project, because this plugin attaches project metadata when it builds Docker images. Simply add this information to any project:

<dependency>
  <groupId>com.spotify</groupId>
  <artifactId>foobar</artifactId>
  <version>1.0-SNAPSHOT</version>
  <type>docker-info</type>
</dependency>

Now, you can read information about the Docker image of the project that you depended on:

String imageName = getResource("META-INF/docker/com.spotify/foobar/image-name");

This is great for an integration test where you want the latest version of another project's Docker image.

Note that you have to register a Maven extension in your POM (or a parent POM) in order for the docker-info type to be supported:

<build>
  <extensions>
    <extension>
      <groupId>com.spotify</groupId>
      <artifactId>dockerfile-maven-extension</artifactId>
      <version>${version}</version>
    </extension>
  </extensions>
</build>

Use other Docker tools that rely on Dockerfiles

Your project(s) look like so:

a/
  Dockerfile
  pom.xml
b/
  Dockerfile
  pom.xml

You can now use these projects with Fig or docker-compose or some other system that works with Dockerfiles. For example, a docker-compose.yml might look like:

service-a:
  build: a/
  ports:
  - '80'

service-b:
  build: b/
  links:
  - service-a

Now, docker-compose up and docker-compose build will work as expected.

Usage

See usage docs.

Authentication

See authentication docs.

Releasing

To cut the Maven release:

mvn clean [-B -Dinvoker.skip -DskipTests -Darguments='-Dinvoker.skip -DskipTests'] \
  -Dgpg.keyname=<key ID used for signing artifacts> \
  release:clean release:prepare release:perform

We use gren to create Releases in Github:

gren release

About

MATURE: A set of Maven tools for dealing with Dockerfiles

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Java 77.1%
  • Groovy 21.0%
  • Dockerfile 1.9%