Adding a new Service Component to GitLab

The GitLab product is made up of several service components that run as independent system processes in communication with each other. These services can be run on the same instance, or spread across different instances. A list of the existing components can be found in the GitLab architecture overview.

Integration phases

The following outline re-uses the maturity metric naming as an example of the various phases of integrating a component. These are only loosely coupled to a components actual maturity, and are intended as a guide for implementation order (for example, a component does not need to be enabled by default to be Lovable, and being enabled by default does not on its own cause a component to be Lovable).

Proposing a new component

The initial step for integrating a new component with GitLab starts with creating a Feature proposal in the issue tracker.

Identify the product category the component falls under and assign the Engineering Manager and Product Manager responsible for that category.

The general steps for getting any GitLab feature from proposal to release can be found in the Product development flow.

Integrating a new service with GitLab

Adding a new service follows the same merge request workflow as other contributions, and must meet the same completion criteria and in addition needs to cover the following:

For services that can operate completely separate from GitLab:

The first iteration should be to add the ability to connect and use the service as an externally installed component. Often this involves providing settings in GitLab to connect to the service, or allow connections from it. And then shipping documentation on how to install and configure the service with GitLab.

Tip: Elasticsearch is an example of a service that has been integrated this way. And many of the other services, including internal projects like Gitaly, started off as separately installed alternatives.

For services that depend on the existing GitLab codebase:

The first iteration should be opt-in, either through the gitlab.yml configuration or through feature flags. For these types of services it is often necessary to bundle the service and its dependencies with GitLab as part of the initial integration.

Tip: ActionCable is an example of a service that has been added this way.

Bundling a service with GitLab

Note: Code shipped with GitLab needs to use a license approved by the Legal team. See the list of existing approved licenses.
Note: Notify the Distribution team when adding a new dependency that must be compiled. We must be able to compile the dependency on all supported platforms.

New services to be bundled with GitLab need to be available in the following environments.

Dev environment

The first step of bundling a new service is to provide it in the development environment to engage in collaboration and feedback.

Standard install methods

In order for a service to be bundled for end-users or GitLab.com, it needs to be included in the standard install methods:

Handling service dependencies

Dependencies should be kept up to date and be tracked for security updates. For the Rails codebase, the JavaScript and Ruby dependencies are scanned for vulnerabilities using GitLab dependency scanning.

In addition, any system dependencies used in Omnibus packages or the Cloud Native images should be added to the dependency update automation.

Release management

If the service component needs to be updated or released with the monthly GitLab release, then the component should be added to the release tools automation. This project is maintained by the Delivery team. A list of the projects managed this way can be found in the release tools project directory.

For example, during the monthly GitLab release, the desired version of Gitaly, GitLab Workhorse, GitLab Shell, etc., need to synchronized through the various release pipelines.