- GitLab Runner versions
- Runner registration
- Configuring runners
- Monitoring runners
- Use a runner to run your job
- Runners on GitLab.com
- Runner execution flow
GitLab Runner is an application that works with GitLab CI/CD to run jobs in a pipeline.
You can choose to install the GitLab Runner application on infrastructure that you own or manage. If you do, you should install GitLab Runner on a machine that’s separate from the one that hosts the GitLab instance for security and performance reasons. When you use separate machines, you can have different operating systems and tools, like Kubernetes or Docker, on each.
GitLab Runner is open-source and written in Go. It can be run as a single binary; no language-specific requirements are needed.
You can install GitLab Runner on several different supported operating systems. Other operating systems may also work, as long as you can compile a Go binary on them.
GitLab Runner can also run inside a Docker container or be deployed into a Kubernetes cluster.
View some best practices for how to use and administer GitLab Runner.
For compatibility reasons, the GitLab Runner major.minor version should stay in sync with the GitLab major and minor version. Older runners may still work with newer GitLab versions, and vice versa. However, features may be not available or work properly if a version difference exists.
Backward compatibility is guaranteed between minor version updates. However, sometimes minor version updates of GitLab can introduce new features that require GitLab Runner to be on the same minor version.
After you install the application, you register individual runners. Runners are the agents that run the CI/CD jobs that come from GitLab.
When you register a runner, you are setting up communication between your GitLab instance and the machine where GitLab Runner is installed.
Runners usually process jobs on the same machine where you installed GitLab Runner. However, you can also have a runner process jobs in a container, in a Kubernetes cluster, or in auto-scaled instances in the cloud.
When you register a runner, you must choose an executor.
An executor determines the environment each job runs in.
- If you want your CI/CD job to run PowerShell commands, you might install GitLab Runner on a Windows server and then register a runner that uses the shell executor.
- If you want your CI/CD job to run commands in a custom Docker container, you might install GitLab Runner on a Linux server and register a runner that uses the Docker executor.
These are only a few of the possible configurations. You can install GitLab Runner on a virtual machine and have it use another virtual machine as an executor.
When you install GitLab Runner in a Docker container and choose the Docker executor to run your jobs, it’s sometimes referred to as a “Docker-in-Docker” configuration.
Before you register a runner, you should determine if everyone in GitLab should have access to it, or if you want to limit it to a specific GitLab group or project.
There are three types of runners, based on who you want to have access:
- Shared runners are for use by all projects
- Group runners are for all projects and subgroups in a group
- Specific runners are for individual projects
When you register a runner, you specify a token for the GitLab instance, group, or project. This is how the runner knows which projects it’s available for.
When you register a runner, you can add tags to it.
When a CI/CD job runs, it knows which runner to use by looking at the assigned tags.
For example, if a runner has the
ruby tag, you would add this code to
job: tags: - ruby
When the job runs, it uses the runner with the
You can configure
the runner by editing the
config.toml file. This is a file that is installed during the runner installation process.
In this file you can edit settings for a specific runner, or for all runners.
You can specify settings like logging and cache. You can set concurrency, memory, CPU limits, and more.
You can use Prometheus to monitor your runners. You can view things like the number of currently-running jobs and how much CPU your runners are using.
After a runner is configured and available for your project, your CI/CD jobs can use the runner.
Specify the name of the runner or its tags in your
Then, when you commit to your repository, the pipeline runs, and
the runner’s executor processes the commands.
If you use GitLab.com, you can run your CI jobs on Runner Cloud. These are runners managed by GitLab and fully integrated with GitLab.com. These runners are enabled for all projects, though you can disable them.
If you don’t want to use runners managed by GitLab, you can install GitLab Runner and register your own runners on GitLab.com.
GitLab Runner has the following features.
- Run multiple jobs concurrently.
- Use multiple tokens with multiple servers (even per-project).
- Limit the number of concurrent jobs per-token.
- Jobs can be run:
- Using Docker containers.
- Using Docker containers and executing job over SSH.
- Using Docker containers with autoscaling on different clouds and virtualization hypervisors.
- Connecting to a remote SSH server.
- Is written in Go and distributed as single binary without any other requirements.
- Supports Bash, PowerShell Core, and Windows PowerShell.
- Works on GNU/Linux, macOS, and Windows (pretty much anywhere you can run Docker).
- Allows customization of the job running environment.
- Automatic configuration reload without restart.
- Easy to use setup with support for Docker, Docker-SSH, Parallels, or SSH running environments.
- Enables caching of Docker containers.
- Easy installation as a service for GNU/Linux, macOS, and Windows.
- Embedded Prometheus metrics HTTP server.
- Referee workers to monitor and pass Prometheus metrics and other job-specific data to GitLab.
Learn how to troubleshoot common issues.
If you’re a reviewer of GitLab Runner project, take a moment to read the Reviewing GitLab Runner document.
You can also review the release process for the GitLab Runner project.
See the CHANGELOG to view recent changes.
This code is distributed under the MIT license. View the LICENSE file.