Day: July 14, 2022

Earlier in my career, our entire development team volunteered to help the reporting team validate some major updates with their service. None of us were excited about pausing our current priorities, but we all realized this was the right thing to do.

After I ran my first report, it appeared that the data connected to the application might be a copy of production data. I was able to quickly validate this assumption by running reports based on my own work-related expenses.

Imagine the following situation: you start working on a new project, maybe also with a different programming language you’re not used to. So, the project is there, and you should be able to run it.

You hope there’s any documentation telling you what to do — which is not that common — and if/when there’s any, it usually doesn’t work. You need to know what to install, where to install it, how to set it up, etc. It’s not an uncommon scenario, and you can actually expect that at some point. But what if there was a way to ensure this won’t happen again?

Throughout this post, we’ll see different approaches we could use to make this easier using only Docker.

Example With Java + Maven

Let’s consider a Java project, for example. Usually, to run a Java application, you run java -jar application.jar.

To generate the jar file and manage project dependencies, you can use a lot of different tools, with the most known being Maven and Gradle. Let’s consider Maven for this example. Let’s see some Maven commands:

• mvn dependency:copy-dependencies
  Downloads the dependencies if they’re not downloaded yet.
• mvn package
  Builds the application and generates the jar. It also downloads the dependencies if they’re not downloaded yet. If you want to skip running the tests in the building process, you can also pass the -Dmaven.test.skip=true parameter.

Assuming we need Maven 3 and Java 11, that’s how you could use Docker:

alias java='docker run -v "$PWD":/home -w /home openjdk:11-jre-slim java'
alias mvn='docker run -it --rm --name maven -v "$(pwd)":/usr/src/mymaven -w /usr/src/mymaven maven:3-jdk-11-slim mvn'

This way, you can run any Maven and Java commands without having to install Java or Maven. You can test the commands by running java -version or mvn -version. Usually, the official Docker image of these tools gives you instructions on how to run it, and you can just create an alias for that.

Pros:

• If you don’t need to use it anymore, you can remove the related Docker image.
• It’s easy to change the version.

Cons:

• In this example, you still need to find out what Java version is used and what tool is used (in this case, Maven) and its version.
• If you’re dealing with a programming language you don’t know, it will take even more time to understand what to do.
• You still need to know which commands to run.

It’s a fair approach, especially if you know what you’re doing. But that doesn’t come with the project itself. So, let’s try to improve it a bit.

That’s where Dockerfile starts to shine. We know how to run commands using only Docker, but how to run the application?

A common Dockerfile for that situation could be:

FROM openjdk:11-jre-slim

ARG JAR_FILE=target/*.jar

ADD ${JAR_FILE} app.jar

ENTRYPOINT ["java", "-jar", "/app.jar"]

And you can build it the way you normally build a docker image, using docker build -t my-application ., for example.

You can see that it depends on an existing JAR file to build it. As you saw on the previous topic, we know how to generate it, but we would be in trouble if it were another programming language or tool.

It seems like a really minor improvement, but that helps a lot already, as you can see in its pros/cons:

Pros

• The Dockerfile should come within the project. So, it tells you already how to run the application regardless of programming language knowledge.
• It also tells you which version and image are being used.
• It inherits the pros from the Level One topic if you also apply Level One knowledge.

Cons

• You still need to find out how to build the application.
• It also means you still need to know which commands to run.

It’s a good approach. You can merge Level One and Level Two to achieve a better result. The project should have Dockerfile, and life gets a bit easier already. But let’s see how life can be even easier.

What if you didn’t know anything about Java and Maven, and you were still able to build the application with just one command you already know?

That’s where Multi-Stage builds shine.

With multi-stage builds, you use multiple FROM statements in your Dockerfile. Each FROM instruction can use a different base, and each of them begins a new stage of the build. You can selectively copy artifacts from one stage to another, leaving behind everything you don’t want in the final image.

How can that help us? Well, let’s consider the previous Dockerfile. We needed a JAR file to build the Docker image. With Multi-Stage Build, Dockerfile itself can be responsible for generating it. In a simple approach, that Dockerfile would look like this:

# ============= DEPENDENCY + BUILD ===========================
# Download the dependencies on container and build application
# ============================================================

FROM maven:3-jdk-11-slim AS builder

COPY ./pom.xml /app/pom.xml
COPY . /app

WORKDIR /app

RUN mvn package $MAVEN_CLI_OPTS -Dmaven.test.skip=true

# ============= DOCKER IMAGE ================
# Prepare container image with application artifacts
# ===========================================

FROM openjdk:11-jre-slim

COPY --from=builder /app/target/*.jar app.jar

ENTRYPOINT ["java", "-jar", "/app.jar"]

Let’s see what’s happening here.

From the first FROM to the first RUN statement, it’s doing things related to Maven — copying the files that need to be copied and running the command that downloads the dependencies and builds the application. It’s doing that by using the maven:3-jdk-11-slim image, and it’s setting the name builder.

After that, you see the second FROM statement using the openjdk:11-jre-slim image. You also see a COPY statement which copies from a place called builder. But what’s that place? What’s that builder?

That’s the name we set for the Maven image at the first FROM statement. It’s copying the jar file from that container. So, you can literally play with different FROM entries to build whatever you want, and the command to build the Docker image is still the same: docker build -t my-application ..

Pros:

• Regardless of the programming language, if the project has this approach, you can run the application without installing anything other than Docker.
• It inherits the pros from Level One and Level Two.

Worth saying that you can also use this Dockerfile with Docker Compose, which can be really powerful, especially if your application needs to expose ports, sharing volumes, or it depends on other images.

Now that you know how to play with different FROM statements, another possible Dockerfile could be:

# ============= DEPENDENCY RESOLVER =============
# Download the dependencies on container
# ===============================================

FROM maven:3-jdk-11-slim AS dependency_resolver

# Download all library dependencies
COPY ./pom.xml /app/pom.xml

WORKDIR /app

RUN mvn dependency:copy-dependencies $MAVEN_CLI_OPTS

# ============= TESTING =================
# Run tests on container
# =======================================

FROM dependency_resolver AS tester

WORKDIR /app

CMD mvn clean test $MAVEN_CLI_OPTS

# ============= BUILDER =================
# Build the artifact on container
# =======================================

FROM dependency_resolver as builder

# Build application
COPY . /app

RUN mvn package $MAVEN_CLI_OPTS -Dmaven.test.skip=true

# ============= DOCKER IMAGE ================
# Prepare container image with application artifacts
# ===========================================

FROM openjdk:11-jre-slim

COPY --from=builder /app/target/*.jar app.jar

ENTRYPOINT ["java", "-jar", "/app.jar"]

So, now we have four different steps: dependency_resolver, tester, builder, and the application itself.

If we want either build the application or test it, we need the project dependencies. Thus, there’s a dependency_resolver step there. In the second and the third FROM statements, you can see that they depend on dependency_resolver.

Important: If you try to build the docker image with docker build -t my-application ., only the first, the third, and the last step (dependency_resolver, builder, and the application itself, respectively) would run.

But why?

When you try to build the image, it will try to see what the end state of the image is, which would be the application itself. As we know and can see, the last step depends on builder (COPY statement). If we check the builder step, we will see that it depends on dependency_resolver (FROM statement). So, it will run in this order:

dependency_resolver -> builder -> application

Then what’s the tester step doing there? Is there a way to reach it?

You can specify a target by using --target: docker build --target tester -t my-application ..

This is also compatible with Docker Compose.

We saw how to use only Docker to build and run applications and commands, removing the need for previous knowledge in some tool or programming language. Also, worth saying that even though we used Docker for these examples, this would also work with other container runtimes like Podman.

I hope you find this post helpful and spread the word about Docker.