How does GitLab Runner work?

GitLab Runner is an agent that executes CI/CD tasks. It exchanges data with GitLab CI/CD via API. When an event occurs in GitLab that triggers a pipeline (for example, a commit), GitLab sends a request to the runner. The runner fetches the task, clones the repository, executes the commands described in .gitlab-ci.yml, and sends the result back to GitLab.

Key aspects of operation:

• Runner types: Shared, Specific, Group. They differ in scope and availability for projects.
• Executor: Defines the environment in which tasks are executed. Popular executors include shell, docker, kubernetes, virtualbox, ssh. The choice of executor affects task isolation and environment.
• Registration: The runner registers with GitLab using a token, linking it to a specific instance, group, or project.
• Polling: The runner periodically polls GitLab for new tasks.
• config.toml: Configuration file defining runner settings, including executor type, number of parallel jobs, and other parameters.

Example structure of config.toml:

# global settings
concurrent = 4

# runner settings
[[runners]]
  name = "my-docker-runner"
  url = "https://gitlab.com/"
  token = "YOUR_RUNNER_TOKEN"
  executor = "docker" # executor choice
  [runners.docker]
    tls_verify = false
    image = "ubuntu:latest" # image when using docker executor
    privileged = false
    disable_entrypoint_overwrite = false
    oom_kill_disable = false
    disable_cache = false
    volumes = ["/cache"]
    shm_size = 0
  [runners.cache]
    [runners.cache.s3]
    [runners.cache.gcs]