TODO:
- : Update README with correct variables names
- : Read SSH Keys / Registration Token from secret file
- : Add a section dedicated to Rancher use
- : Describe volume to mount for persistence
A customized Docker image for running scalable GitLab CI runners on DC/OS via Marathon.
The GitLab runner can be configured by environment variables. For a complete overview, have a look at the docs/gitlab_runner_register_arguments.md file.
The most important ones are:
GITLAB_SERVICE_NAME
: The Mesos DNS service name of the GitLab instance, e.g.gitlab.marathon.mesos
. This strongly depends on your setup, i.e. how you launched GitLab and how you configured Mesos DNS. This is the recommended method to use with DC/OS installations of GitLab. Either this environment variable orGITLAB_INSTANCE_URL
is mandatory.GITLAB_INSTANCE_URL
: The URL of the GitLab instance to connect to, e.g.http://gitlab.mycompany.com
. Either this environment variable orGITLAB_SERVICE_NAME
is mandatory.REGISTRATION_TOKEN
: The registration token to use with the GitLab instance. See the docs for details. (mandatory)RUNNER_EXECUTOR
: The type of the executor to use, e.g.shell
ordocker
. See the executor docs for more details. (mandatory)RUNNER_CONCURRENT_BUILDS
: The number of concurrent builds this runner should be able to handel. Default is1
.RUNNER_TAG_LIST
: If you want to use tags in you.gitlab-ci.yml
, then you need to specify the comma-separated list of tags. This is useful to distinguish the runner types.
Private Docker registries can be used by adding the secret variable DOCKER_AUTH_CONFIG
to your project's Settings ➔ CI/CD Pipelines settings. Have a look at the guide as well.
This version of the GitLab CI runner for Marathon project uses Docker-in-Docker techniques, with all of its pros and cons. See also jpetazzo's article on this topic.
In the following examples, we assume that you're running the GitLab Universe package as service gitlab
on the DC/OS internal Marathon instance, which is also available to the runners via the external_url
of the GitLab configuration. This normally means that GitLab is exposed on a public agent node via marathon-lb. Please see the [example documentation here|https://github.com/dcos/examples/tree/master/1.8/gitlab].
An example for a shell runner. This enables the build of Docker images.
{
"id": "gitlab-runner-shell",
"container": {
"type": "DOCKER",
"docker": {
"image": "mesosphere/dcos-gitlab-runner-service:v9.1.1",
"network": "HOST",
"forcePullImage": true,
"privileged": true
}
},
"instances": 1,
"cpus": 1,
"mem": 2048,
"env": {
"GITLAB_SERVICE_NAME": "gitlab.marathon.mesos",
"REGISTRATION_TOKEN": "zzNWmRE--SBfeMfiKCMh",
"RUNNER_EXECUTOR": "shell",
"RUNNER_TAG_LIST": "shell,build-as-docker",
"RUNNER_CONCURRENT_BUILDS": "4"
},
"taskKillGracePeriodSeconds": 15,
"healthChecks": [
{
"path": "/metrics",
"portIndex": 0,
"protocol": "HTTP",
"gracePeriodSeconds": 300,
"intervalSeconds": 60,
"timeoutSeconds": 20,
"maxConsecutiveFailures": 3,
"ignoreHttp1xx": false
}
]
}
Here's an example for a Docker runner, which enables builds inside Docker containers:
{
"id": "gitlab-runner-docker",
"container": {
"type": "DOCKER",
"docker": {
"image": "mesosphere/dcos-gitlab-runner-service:v9.1.1",
"network": "HOST",
"forcePullImage": true,
"privileged": true
}
},
"instances": 1,
"cpus": 1,
"mem": 2048,
"env": {
"GITLAB_SERVICE_NAME": "gitlab.marathon.mesos",
"REGISTRATION_TOKEN": "zzNWmRE--SBfeMfiKCMh",
"RUNNER_EXECUTOR": "docker",
"RUNNER_TAG_LIST": "docker,build-in-docker",
"RUNNER_CONCURRENT_BUILDS": "4",
"DOCKER_IMAGE": "node:6-wheezy"
},
"taskKillGracePeriodSeconds": 15,
"healthChecks": [
{
"path": "/metrics",
"portIndex": 0,
"protocol": "HTTP",
"gracePeriodSeconds": 300,
"intervalSeconds": 60,
"timeoutSeconds": 20,
"maxConsecutiveFailures": 3,
"ignoreHttp1xx": false
}
]
}
Make sure you choose a useful default Docker image via DOCKER_IMAGE
, for example if you want to build Node.js projects, the node:6-wheezy
image. This can be overwritten with the image
property in the .gitlab-ci.yml
file (see the GitLab CI docs.
An .gitlab-ci.yml
example of using the build-as-docker
tag to trigger a build on the runner(s) with shell executors:
stages:
- ci
build-job:
stage: ci
tags:
- build-as-docker
script:
- docker build -t tobilg/test .
This assumes your project has a Dockerfile
, for example
FROM nginx
An .gitlab-ci.yml
example of using the build-in-docker
tag to trigger a build on the runner(s) with Docker executors:
image: node:6-wheezy
stages:
- ci
test-job:
stage: ci
tags:
- build-in-docker
script:
- node --version