Skip to content

Building Docker images with GitLab CI/CD

GitLab CI/CD allows you to use Docker Engine to build and test docker-based projects.

TIP: Tip: This also allows to you to use docker-compose and other docker-enabled tools.

One of the new trends in Continuous Integration/Deployment is to:

  1. Create an application image
  2. Run tests against the created image
  3. Push image to a remote registry
  4. Deploy to a server from the pushed image

It's also useful when your application already has the Dockerfile that can be used to create and test an image:

docker build -t my-image dockerfiles/
docker run my-docker-image /script/to/run/tests
docker tag my-image my-registry:5000/my-image
docker push my-registry:5000/my-image

This requires special configuration of GitLab Runner to enable docker support during jobs.

Runner Configuration

There are three methods to enable the use of docker build and docker run during jobs; each with their own tradeoffs.

Use shell executor

The simplest approach is to install GitLab Runner in shell execution mode. GitLab Runner then executes job scripts as the gitlab-runner user.

  1. Install GitLab Runner.

  2. During GitLab Runner installation select shell as method of executing job scripts or use command:

    sudo gitlab-runner register -n \
  --url https://gitlab.com/ \
  --registration-token REGISTRATION_TOKEN \
  --executor shell \
  --description "My Runner"

  3. Install Docker Engine on server.

    For more information how to install Docker Engine on different systems checkout the Supported installations.

  4. Add gitlab-runner user to docker group:

    sudo usermod -aG docker gitlab-runner

  5. Verify that gitlab-runner has access to Docker:

    sudo -u gitlab-runner -H docker info

    You can now verify that everything works by adding docker info to .gitlab-ci.yml:

    before_script:
  - docker info

build_image:
  script:
    - docker build -t my-docker-image .
    - docker run my-docker-image /script/to/run/tests

  6. You can now use docker command and install docker-compose if needed.

NOTE: Note: By adding gitlab-runner to the docker group you are effectively granting gitlab-runner full root permissions. For more information please read On Docker security: docker group considered harmful.

Use docker-in-docker executor

The second approach is to use the special docker-in-docker (dind) Docker image with all tools installed (docker and docker-compose) and run the job script in context of that image in privileged mode.

In order to do that, follow the steps:

  1. Install GitLab Runner.

  2. Register GitLab Runner from the command line to use docker and privileged mode:

    sudo gitlab-runner register -n \
  --url https://gitlab.com/ \
  --registration-token REGISTRATION_TOKEN \
  --executor docker \
  --description "My Docker Runner" \
  --docker-image "docker:stable" \
  --docker-privileged

    The above command will register a new Runner to use the special docker:stable image which is provided by Docker. Notice that it's using the privileged mode to start the build and service containers. If you want to use docker-in-docker mode, you always have to use privileged = true in your Docker containers.

    The above command will create a config.toml entry similar to this:

    [[runners]]
  url = "https://gitlab.com/"
  token = TOKEN
  executor = "docker"
  [runners.docker]
    tls_verify = false
    image = "docker:stable"
    privileged = true
    disable_cache = false
    volumes = ["/cache"]
  [runners.cache]
    Insecure = false

  3. You can now use docker in the build script (note the inclusion of the docker:dind service):

    image: docker:stable

variables:
  # When using dind service we need to instruct docker, to talk with the
  # daemon started inside of the service. The daemon is available with
  # a network connection instead of the default /var/run/docker.sock socket.
  #
  # The 'docker' hostname is the alias of the service container as described at
  # https://docs.gitlab.com/ee/ci/docker/using_docker_images.html#accessing-the-services
  #
  # Note that if you're using Kubernetes executor, the variable should be set to
  # tcp://localhost:2375 because of how Kubernetes executor connects services
  # to the job container
  DOCKER_HOST: tcp://docker:2375/
  # When using dind, it's wise to use the overlayfs driver for
  # improved performance.
  DOCKER_DRIVER: overlay2

services:
- docker:dind

before_script:
- docker info

build:
  stage: build
  script:
  - docker build -t my-docker-image .
  - docker run my-docker-image /script/to/run/tests

Docker-in-Docker works well, and is the recommended configuration, but it is not without its own challenges:

  • By enabling --docker-privileged, you are effectively disabling all of the security mechanisms of containers and exposing your host to privilege escalation which can lead to container breakout. For more information, check out the official Docker documentation on Runtime privilege and Linux capabilities.

  • When using docker-in-docker, each job is in a clean environment without the past history. Concurrent jobs work fine because every build gets it's own instance of Docker engine so they won't conflict with each other. But this also means jobs can be slower because there's no caching of layers.

  • By default, docker:dind uses --storage-driver vfs which is the slowest form offered. To use a different driver, see Using the overlayfs driver.

  • Since the docker:dind container and the runner container don't share their root filesystem, the job's working directory can be used as a mount point for children containers. For example, if you have files you want to share with a child container, you may create a subdirectory under /builds/$CI_PROJECT_PATH and use it as your mount point (for a more thorough explanation, check issue #41227):

    variables:
  MOUNT_POINT: /builds/$CI_PROJECT_PATH/mnt

script:
  - mkdir -p "$MOUNT_POINT"
  - docker run -v "$MOUNT_POINT:/mnt" my-docker-image

An example project using this approach can be found here: https://gitlab.com/gitlab-examples/docker.

Use Docker socket binding

The third approach is to bind-mount /var/run/docker.sock into the container so that docker is available in the context of that image.

In order to do that, follow the steps:

  1. Install GitLab Runner.

  2. Register GitLab Runner from the command line to use docker and share /var/run/docker.sock:

    sudo gitlab-runner register -n \
  --url https://gitlab.com/ \
  --registration-token REGISTRATION_TOKEN \
  --executor docker \
  --description "My Docker Runner" \
  --docker-image "docker:stable" \
  --docker-volumes /var/run/docker.sock:/var/run/docker.sock

    The above command will register a new Runner to use the special docker:stable image which is provided by Docker. Notice that it's using the Docker daemon of the Runner itself, and any containers spawned by docker commands will be siblings of the Runner rather than children of the runner. This may have complications and limitations that are unsuitable for your workflow.

    The above command will create a config.toml entry similar to this:

    [[runners]]
  url = "https://gitlab.com/"
  token = REGISTRATION_TOKEN
  executor = "docker"
  [runners.docker]
    tls_verify = false
    image = "docker:stable"
    privileged = false
    disable_cache = false
    volumes = ["/var/run/docker.sock:/var/run/docker.sock", "/cache"]
  [runners.cache]
    Insecure = false

  3. You can now use docker in the build script (note that you don't need to include the docker:dind service as when using the Docker in Docker executor):

    image: docker:stable

before_script:
- docker info

build:
  stage: build
  script:
  - docker build -t my-docker-image .
  - docker run my-docker-image /script/to/run/tests

While the above method avoids using Docker in privileged mode, you should be aware of the following implications:

  • By sharing the docker daemon, you are effectively disabling all the security mechanisms of containers and exposing your host to privilege escalation which can lead to container breakout. For example, if a project ran docker rm -f $(docker ps -a -q) it would remove the GitLab Runner containers.

  • Concurrent jobs may not work; if your tests create containers with specific names, they may conflict with each other.

  • Sharing files and directories from the source repo into containers may not work as expected since volume mounting is done in the context of the host machine, not the build container, e.g.:

    docker run --rm -t -i -v $(pwd)/src:/home/app/src test-image:latest run_app_tests

Making docker-in-docker builds faster with Docker layer caching

When using docker-in-docker, Docker will download all layers of your image every time you create a build. Recent versions of Docker (Docker 1.13 and above) can use a pre-existing image as a cache during the docker build step, considerably speeding up the build process.

How Docker caching works

When running docker build, each command in Dockerfile results in a layer. These layers are kept around as a cache and can be reused if there haven't been any changes. Change in one layer causes all subsequent layers to be recreated.

You can specify a tagged image to be used as a cache source for the docker build command by using the --cache-from argument. Multiple images can be specified as a cache source by using multiple --cache-from arguments. Keep in mind that any image that's used with the --cache-from argument must first be pulled (using docker pull) before it can be used as a cache source.

Using Docker caching

Here's a simple .gitlab-ci.yml file showing how Docker caching can be utilized:

image: docker:stable

services:
  - docker:dind

variables:
  CONTAINER_IMAGE: registry.gitlab.com/$CI_PROJECT_PATH
  DOCKER_HOST: tcp://docker:2375
  DOCKER_DRIVER: overlay2

before_script:
  - docker login -u gitlab-ci-token -p $CI_JOB_TOKEN registry.gitlab.com

build:
  stage: build
  script:
    - docker pull $CONTAINER_IMAGE:latest || true
    - docker build --cache-from $CONTAINER_IMAGE:latest --tag $CONTAINER_IMAGE:$CI_BUILD_REF --tag $CONTAINER_IMAGE:latest .
    - docker push $CONTAINER_IMAGE:$CI_BUILD_REF
    - docker push $CONTAINER_IMAGE:latest

The steps in the script section for the build stage can be summed up to:

  1. The first command tries to pull the image from the registry so that it can be used as a cache for the docker build command.
  2. The second command builds a Docker image using the pulled image as a cache (notice the --cache-from $CONTAINER_IMAGE:latest argument) if available, and tags it.
  3. The last two commands push the tagged Docker images to the container registry so that they may also be used as cache for subsequent builds.

Using the OverlayFS driver

NOTE: Note: The shared Runners on GitLab.com use the overlay2 driver by default.

By default, when using docker:dind, Docker uses the vfs storage driver which copies the filesystem on every run. This is a very disk-intensive operation which can be avoided if a different driver is used, for example overlay2.

Requirements

  1. Make sure a recent kernel is used, preferably >= 4.2.

  2. Check whether the overlay module is loaded:

    sudo lsmod | grep overlay

    If you see no result, then it isn't loaded. To load it use:

    sudo modprobe overlay

    If everything went fine, you need to make sure module is loaded on reboot. On Ubuntu systems, this is done by editing /etc/modules. Just add the following line into it:

    overlay

Use driver per project

You can enable the driver for each project individually by editing the project's .gitlab-ci.yml:

variables:
  DOCKER_DRIVER: overlay2

Use driver for every project

To enable the driver for every project, you can set the environment variable for every build by adding environment in the [[runners]] section of config.toml:

environment = ["DOCKER_DRIVER=overlay2"]

If you're running multiple Runners you will have to modify all configuration files.

Notes:

Using the GitLab Container Registry

Notes:

  • This feature requires GitLab 8.8 and GitLab Runner 1.2.
  • Starting from GitLab 8.12, if you have 2FA enabled in your account, you need to pass a personal access token instead of your password in order to login to GitLab's Container Registry.

Once you've built a Docker image, you can push it up to the built-in GitLab Container Registry. For example, if you're using docker-in-docker on your runners, this is how your .gitlab-ci.yml could look like:

 build:
   image: docker:stable
   services:
   - docker:dind
   variables:
     DOCKER_HOST: tcp://docker:2375
     DOCKER_DRIVER: overlay2
   stage: build
   script:
     - docker login -u gitlab-ci-token -p $CI_JOB_TOKEN registry.example.com
     - docker build -t registry.example.com/group/project/image:latest .
     - docker push registry.example.com/group/project/image:latest

You have to use the special gitlab-ci-token user created for you in order to push to the Registry connected to your project. Its password is provided in the $CI_JOB_TOKEN variable. This allows you to automate building and deployment of your Docker images.

You can also make use of other variables to avoid hardcoding:

services:
  - docker:dind

variables:
  DOCKER_HOST: tcp://docker:2375
  DOCKER_DRIVER: overlay2
  IMAGE_TAG: $CI_REGISTRY_IMAGE:$CI_COMMIT_REF_NAME

before_script:
  - docker login -u gitlab-ci-token -p $CI_JOB_TOKEN $CI_REGISTRY

build:
  stage: build
  script:
    - docker build -t $IMAGE_TAG .
    - docker push $IMAGE_TAG

Here, $CI_REGISTRY_IMAGE would be resolved to the address of the registry tied to this project, and $CI_COMMIT_REF_NAME would be resolved to the branch or tag name for this particular job. We also declare our own variable, $IMAGE_TAG, combining the two to save us some typing in the script section.

Here's a more elaborate example that splits up the tasks into 4 pipeline stages, including two tests that run in parallel. The build is stored in the container registry and used by subsequent stages, downloading the image when needed. Changes to master also get tagged as latest and deployed using an application-specific deploy script:

image: docker:stable
services:
- docker:dind

stages:
- build
- test
- release
- deploy

variables:
  DOCKER_HOST: tcp://docker:2375
  DOCKER_DRIVER: overlay2
  CONTAINER_TEST_IMAGE: registry.example.com/my-group/my-project/my-image:$CI_COMMIT_REF_NAME
  CONTAINER_RELEASE_IMAGE: registry.example.com/my-group/my-project/my-image:latest

before_script:
  - docker login -u gitlab-ci-token -p $CI_JOB_TOKEN registry.example.com

build:
  stage: build
  script:
    - docker build --pull -t $CONTAINER_TEST_IMAGE .
    - docker push $CONTAINER_TEST_IMAGE

test1:
  stage: test
  script:
    - docker pull $CONTAINER_TEST_IMAGE
    - docker run $CONTAINER_TEST_IMAGE /script/to/run/tests

test2:
  stage: test
  script:
    - docker pull $CONTAINER_TEST_IMAGE
    - docker run $CONTAINER_TEST_IMAGE /script/to/run/another/test

release-image:
  stage: release
  script:
    - docker pull $CONTAINER_TEST_IMAGE
    - docker tag $CONTAINER_TEST_IMAGE $CONTAINER_RELEASE_IMAGE
    - docker push $CONTAINER_RELEASE_IMAGE
  only:
    - master

deploy:
  stage: deploy
  script:
    - ./deploy.sh
  only:
    - master

Some things you should be aware of when using the Container Registry:

  • You must log in to the container registry before running commands. Putting this in before_script will run it before each job.
  • Using docker build --pull makes sure that Docker fetches any changes to base images before building just in case your cache is stale. It takes slightly longer, but means you don’t get stuck without security patches to base images.
  • Doing an explicit docker pull before each docker run makes sure to fetch the latest image that was just built. This is especially important if you are using multiple runners that cache images locally. Using the git SHA in your image tag makes this less necessary since each job will be unique and you shouldn't ever have a stale image, but it's still possible if you re-build a given commit after a dependency has changed.
  • You don't want to build directly to latest in case there are multiple jobs happening simultaneously.