moby/docs/contributing/set-up-dev-env.md

Work with a development container

In this section, you learn to develop like the Moby Engine core team. The moby/moby repository includes a Dockerfile at its root. This file defines Moby's development environment. The Dockerfile lists the environment's dependencies: system libraries and binaries, Go environment, Go dependencies, etc.

Moby's development environment is itself, ultimately a Docker container. You use the moby/moby repository and its Dockerfile to create a Docker image, run a Docker container, and develop code in the container.

If you followed the procedures that set up Git for contributing, you should have a fork of the moby/moby repository. You also created a branch called dry-run-test. In this section, you continue working with your fork on this branch.

Task 1. Remove images and containers

Moby developers run the latest stable release of the Docker software. They clean their local hosts of unnecessary Docker artifacts such as stopped containers or unused images. Cleaning unnecessary artifacts isn't strictly necessary, but it is good practice, so it is included here.

To remove unnecessary artifacts:

1. Verify that you have no unnecessary containers running on your host.

   $ docker ps -a
   

   You should see something similar to the following:

   CONTAINER ID        IMAGE               COMMAND             CREATED             STATUS              PORTS               NAMES
   

   There are no running or stopped containers on this host. A fast way to remove old containers is the following:

   You can now use the docker system prune command to achieve this:

   $ docker system prune -a
   

   Older versions of the Docker Engine should reference the command below:

   $ docker rm $(docker ps -a -q)
   

   This command uses docker ps to list all containers (-a flag) by numeric IDs (-q flag). Then, the docker rm command removes the resulting list. If you have running but unused containers, stop and then remove them with the docker stop and docker rm commands.

2. Verify that your host has no dangling images.

   $ docker images
   

   You should see something similar to the following:

   REPOSITORY          TAG                 IMAGE ID            CREATED             SIZE
   

   This host has no images. You may have one or more dangling images. A dangling image is not used by a running container and is not an ancestor of another image on your system. A fast way to remove dangling image is the following:

   $ docker rmi -f $(docker images -q -a -f dangling=true)
   

   This command uses docker images to list all images (-a flag) by numeric IDs (-q flag) and filter them to find dangling images (-f dangling=true). Then, the docker rmi command forcibly (-f flag) removes the resulting list. If you get a "docker: "rmi" requires a minimum of 1 argument." message, that means there were no dangling images. To remove just one image, use the docker rmi ID command.

Task 2. Start a development container

If you followed the last procedure, your host is clean of unnecessary images and containers. In this section, you build an image from the Engine development environment and run it in the container. Both steps are automated for you by the Makefile in the Engine code repository. The first time you build an image, it can take over 15 minutes to complete.

1. Open a terminal.

   For Docker Toolbox users, use docker-machine status your_vm_name to make sure your VM is running. You may need to run eval "$(docker-machine env your_vm_name)" to initialize your shell environment. If you use Docker for Mac or Docker for Windows, you do not need to use Docker Machine.

2. Change into the root of the moby-fork repository.

   $ cd ~/repos/moby-fork
   

   If you are following along with this guide, you created a dry-run-test branch when you set up Git for contributing.

3. Ensure you are on your dry-run-test branch.

   $ git checkout dry-run-test
   

   If you get a message that the branch doesn't exist, add the -b flag (git checkout -b dry-run-test) so the command both creates the branch and checks it out.

4. Use make to build a development environment image and run it in a container.

   $ make shell
   

   Using the instructions in the Dockerfile, the build may need to download and / or configure source and other images. On first build this process may take between 5 - 15 minutes to create an image. The command returns informational messages as it runs. A successful build returns a final message and opens a Bash shell into the container.

   Successfully built 3d872560918e
   docker run --rm -i --privileged -e BUILDFLAGS -e KEEPBUNDLE -e DOCKER_CLIENTONLY -e DOCKER_DEBUG -e DOCKER_EXPERIMENTAL -e DOCKER_GITCOMMIT -e DOCKER_GRAPHDRIVER=vfs -e DOCKER_REMAP_ROOT -e DOCKER_STORAGE_OPTS -e DOCKER_USERLANDPROXY -e TESTDIRS -e TESTFLAGS -e TIMEOUT -v "home/ubuntu/repos/docker/bundles:/go/src/github.com/docker/docker/bundles" -t "docker-dev" bash
   #
   

   At this point, your prompt reflects the container's BASH shell.

   Alternatively you can use the provided devcontainer in an IDE that supports them (VSCode, Goland, etc.)

Note: The make shell command creates an image tagged as docker-dev:latest by default. It does not automatically tag the image with your current branch name, even if you are on a feature branch. Some older documentation or examples may refer to a branch-specific tag, but that behavior is no longer used.

5. List the contents of the current directory (/go/src/github.com/docker/docker).

   You should see the image's source from the /go/src/github.com/docker/docker directory.

   

6. Make a dockerd binary.

   # hack/make.sh binary
   Removing bundles/
   
   ---> Making bundle: binary (in bundles/binary)
   Building bundles/binary-daemon/dockerd (linux/amd64)...
   Created binary: bundles/binary-daemon/dockerd
   Building bundles/binary-daemon/docker-proxy (linux/amd64)...
   Created binary:bundles/binary-daemon/docker-proxy
   

7. Run make install, which copies the binary to the container's /usr/local/bin/ directory.

   # make install
   

8. Start the Engine daemon running in the background.

   # dockerd -D &
   ...output snipped...
   DEBU[0001] Registering POST, /networks/{id:.*}/connect
   DEBU[0001] Registering POST, /networks/{id:.*}/disconnect
   DEBU[0001] Registering DELETE, /networks/{id:.*}
   INFO[0001] API listen on /var/run/docker.sock
   DEBU[0003] containerd connection state change: READY
   

   The -D flag starts the daemon in debug mode. The & starts it as a background process. You'll find these options useful when debugging code development. You will need to hit return in order to get back to your shell prompt.

   Note

   : The following command automates the build, install, and run steps above. Once the command below completes, hit ctrl-z to suspend the process, then run bg 1 and hit enter to resume the daemon process in the background and get back to your shell prompt.

   hack/make.sh binary install-binary run
   

9. Inside your container, check your Docker versions:

   # docker version
   Client:
    Version:      17.06.0-ce
    API version:  1.30
    Go version:   go1.8.3
    Git commit:   02c1d87
    Built:        Fri Jun 23 21:15:15 2017
    OS/Arch:      linux/amd64
   
   Server:
    Version:      dev
    API version:  1.35 (minimum version 1.12)
    Go version:   go1.9.2
    Git commit:   4aa6362da
    Built:        Sat Dec  2 05:22:42 2017
    OS/Arch:      linux/amd64
    Experimental: false
   

   Notice the split versions between client and server, which might be unexpected. In more recent times the Docker CLI component (which provides the docker command) has split out from the Moby project and is now maintained in docker/cli.

   The Moby project now defaults to a fixed version of the docker CLI for integration tests.

   You may have noticed the following message when starting the container with the shell command:

   Makefile:123: The docker client CLI has moved to github.com/docker/cli. For a dev-test cycle involving the CLI, run:
   DOCKER_CLI_PATH=/host/path/to/cli/binary make shell
   then change the cli and compile into a binary at the same location.
   

   By setting DOCKER_CLI_PATH you can supply a newer docker CLI to the server development container for testing and for integration-cli test-execution:

   make DOCKER_CLI_PATH=~/go/src/github.com/docker/cli/build/docker shell
   ...
   # which docker
   /usr/local/cli/docker
   # docker --version
   Docker version 29.0.0-dev, build 09cd4ea26c
   

   This Docker CLI should be built from the docker-cli project and needs to be a Linux binary.

   Inside the container you are running a development version. This is the version on the current branch. It reflects the value of the VERSION file at the root of your docker-fork repository.

10. Run the hello-world image.

    # docker run hello-world
    

11. List the image you just downloaded.

    # docker images
    REPOSITORY   TAG     IMAGE ID      CREATED        SIZE
    hello-world  latest  c54a2cc56cbb  3 months ago   1.85 kB
    

12. Open another terminal on your local host.

13. List the container running your development container.

    ubuntu@ubuntu1404:~$ docker ps
    CONTAINER ID   IMAGE        COMMAND            CREATED          STATUS          PORTS     NAMES
    a8b2885ab900   docker-dev   "hack/dind bash"   43 minutes ago   Up 43 minutes             hungry_payne
    

Task 3. Make a code change

At this point, you have experienced the "Moby inception" technique. That is, you have:

• forked and cloned the Moby Engine code repository
• created a feature branch for development
• created and started an Engine development container from your branch
• built a binary inside your development container
• launched a docker daemon using your newly compiled binary
• called the docker client to run a hello-world container inside your development container

Running the make shell command mounted your local Docker repository source into your Docker container.

When you start to develop code though, you'll want to iterate code changes and builds inside the container. If you have followed this guide exactly, you have a bash shell running a development container.

Try a simple code change and see it reflected in your container. For this example, you'll edit the help for the attach subcommand.

1. If you don't have one, open a terminal in your local host.

2. Make sure you are in your moby-fork repository.

   $ pwd
   /Users/mary/go/src/github.com/moxiegirl/moby-fork
   

   Your location should be different because, at least, your username is different.

3. Open the daemon/command/docker.go file.

4. Edit the command's help message.

   For example, you can edit this line:

   Short:         "A self-sufficient runtime for containers.",
   

   And change it to this:

   Short:         "A self-sufficient and really fun runtime for containers.",
   

5. Save and close the daemon/command/docker.go file.

6. Go to your running docker development container shell.

7. Rebuild the binary by using the command hack/make.sh binary in the docker development container shell.

8. Stop Docker if it is running.

9. Copy the binaries to /usr/local/bin by entering the following commands in the docker development container shell.

   hack/make.sh binary install-binary
   

10. To view your change, run the dockerd --help command in the docker development container shell.

# dockerd --help

Usage:        dockerd COMMAND

A self-sufficient and really fun runtime for containers.

Options:
...

You've just done the basic workflow for changing the Engine code base. You made your code changes in your feature branch. Then, you updated the binary in your development container and tried your change out. If you were making a bigger change, you might repeat or iterate through this flow several times.

Where to go next

Congratulations, you have successfully achieved Docker inception. You've had a small experience of the development process. You've set up your development environment and verified almost all the essential processes you need to contribute. Of course, before you start contributing, you'll need to learn one more piece of the development process, the test framework.