Fully dockerized Gitlab Runner
This image was forked from rayrutjes/simple-gitlab-runner. It was updated to reflect the current
gitlab-ci-multi-runner command syntax, as well as to contain some further instructions.
It registers and runs a single gitlab runner. If you need multiple runners, either supply the
CONCURRENT environment variable (with an integer value, e.g. 4), or simply start multiple containers.
The usage examples below focus on running this container with the necessary configuration to let you
spawn docker containers from inside it.
Note that we share the docker.sock instead of using some dind image. please read this blog post to understand our motivations to avoid Docker in Docker strategy.
Be aware that images are built in the host. This is great because it allows you to share the images cache and run your builds faster. Take care to you always include dynamic tags to your inner docker builds so that parallel builds don't conflict.
Also take note that if you use docker-compose in your tests, you would need to ensure that concurrent builds don't conflict. Actually, gitlab ci automatically puts your current build in a folder with the name of your project. Docker-compose will use this folder name to name your images and containers. If you run multiple containers of this image on the same host and you want to use docker-compose, you will need to find a way to isolate the builds. Please feel free to contribute on the subject if you have an idea.
Manual usage example
docker run -d --name gitlab-runner \ -v /usr/bin/docker:/usr/bin/docker \ -v /var/run/docker.sock:/var/run/docker.sock \ -v /opt/gitlab-runner:/opt/gitlab-runner \ tobilg/gitlab-runner \ -u 'YOUR_CI_URL' \ -r 'YOUR_REGISTRATION_TOKEN' \ --name 'YOUR_RUNNER_NAME' \ --tag-list 'YOUR_COMMA_SEPARATED_TAGS' \ --executor 'shell'
All the command arguments are proxied to the
gitlab-runner register -n. For more information on available options check the official documentation of the gitlab runner.
Mount the docker socket so that builds can spawn docker containers through the host's docker.
Mount the docker binary so that the versions match. Adapt this if the docker resides in another directory.
Mount the work directory. The directory must be the same on the host and inside the container, so that new containers can mount volumes from the current build using
pwd for example.
If your host system is CoreOS, you'll have to add
to the Docker command. Otherwise Docker will not work in the container.
We actually do no recommend mounting the config directory /etc/gitlab-runner as this container is intended to be ephemeral and run a single runner.