- Selecting the executor
- Compatibility chart
GitLab Runner implements a number of executors that can be used to run your builds in different scenarios. If you are not sure what to select, read the I am not sure section. Visit the compatibility chart to find out what features each executor does and does not support.
To jump into the specific documentation for each executor, visit:
The list of executors above is locked. We no longer are developing or accepting new ones. Please check CONTRIBUTION.md to check the details.
The executors support different platforms and methodologies for building a project. The table below shows the key facts for each executor which will help you decide which executor to use.
|Clean build environment for every build||✗||✗||✓||✓||✓||✓||conditional (4)|
|Reuse previous clone if it exists||✓||✓||✗||✗||✓||✗||conditional (4)|
|Migrate runner machine||✗||✗||partial||partial||✓||✓||✓|
|Zero-configuration support for concurrent builds||✗||✗ (1)||✓||✓||✓||✓||conditional (4)|
|Complicated build environments||✗||✗ (2)||✓ (3)||✓ (3)||✓||✓||✓|
|Debugging build problems||easy||easy||hard||hard||medium||medium||medium|
- It’s possible, but in most cases it is problematic if the build uses services installed on the build machine
- It requires to install all dependencies by hand
- For example using Vagrant
- Dependent on what kind of environment you are provisioning. It can be completely isolated or shared between each build.
Shell is the simplest executor to configure. All required dependencies for your builds need to be installed manually on the same machine that the Runner is installed on.
This type of executor allows you to use an already created virtual machine, which is cloned and used to run your build. We offer two full system virtualization options: VirtualBox and Parallels. They can prove useful if you want to run your builds on different operating systems, since it allows the creation of virtual machines on Windows, Linux, macOS or FreeBSD, then GitLab Runner connects to the virtual machine and runs the build on it. Its usage can also be useful for reducing infrastructure costs.
A great option is to use Docker as it allows a clean build environment, with easy dependency management (all dependencies for building the project can be put in the Docker image). The Docker executor allows you to easily create a build environment with dependent services, like MySQL.
The Docker Machine is a special version of the Docker executor with support for auto-scaling. It works like the normal Docker executor but with build hosts created on demand by Docker Machine.
The Kubernetes executor allows you to use an existing Kubernetes cluster for your builds. The executor will call the Kubernetes cluster API and create a new Pod (with a build container and services containers) for each GitLab CI job.
The SSH executor is added for completeness, but it’s the least supported among all executors. It makes GitLab Runner connect to an external server and runs the builds there. We have some success stories from organizations using this executor, but usually we recommend using one of the other types.
The Custom executor allows you to specify your own execution environments. When GitLab Runner does not provide an executor (for example, LXC containers), you are able to provide your own executables to GitLab Runner to provision and clean up any environment you want to use.
Supported features by different executors:
|GitLab Runner Exec command||✗||✓||✗||✗||✓||✓||✓|
|Passing artifacts between stages||✓||✓||✓||✓||✓||✓||✓|
|Use GitLab Container Registry private images||n/a||n/a||n/a||n/a||✓||✓||n/a|
|Interactive Web terminal||✗||✓ (bash)||✗||✗||✓||✓||✗|
Supported systems by different shells:
|Windows||✗ (4)||✓ (2)||✓ (3)|
- Default shell.
- Default shell if no
- Default shell when a new GitLab Runner is registered.
- Bash shell is currently not working on Windows out of the box due to this issue but is intended to be supported again soon. See the issue for a workaround.
Supported systems for interactive web terminals by different shells: