- Components repository
- Create a components repository
- Release a component
- Use a component in a CI/CD configuration
- Best practices
- Convert a CI/CD template to a component
CI/CD components
- Introduced as an experimental feature in GitLab 16.0, with a flag named
ci_namespace_catalog_experimental
. Disabled by default. - Enabled on GitLab.com and self-managed in GitLab 16.2.
-
Feature flag
ci_namespace_catalog_experimental
removed in GitLab 16.3.
This feature is an experimental feature and an epic exists to track future work. Tell us about your use case by leaving comments in the epic.
Components repository
A components repository is a GitLab project with a repository that hosts one or more pipeline components. A pipeline component is a reusable single pipeline configuration unit. Use them to compose an entire pipeline configuration or a small part of a larger pipeline.
A component can optionally take input parameters.
Create a components repository
To create a components repository, you must:
-
Create a new project with a
README.md
file. -
Create a
template.yml
file inside the project’s root directory that contains the configuration you want to provide as a component. For example:spec: inputs: stage: default: test --- component-job: script: echo job 1 stage: $[[ inputs.stage ]]
Directory structure
A components repository can host one or more components, and must follow a mandatory file structure.
Component configurations can be saved through the following directory structure, containing:
- A
templates
directory at the top level of your components repository. All component configuration files should be saved under this directory. - Files ending in
.yml
containing the component configurations, one file per component. - A Markdown
README.md
file explaining the details of all the components in the repository.
For example, if the project contains a single component and a pipeline to test the component, the file structure should be similar to:
├── templates/
│ └── only_template.yml
├── README.md
└── .gitlab-ci.yml
This example component could be referenced with a path similar to gitlab.com/my-username/my-component/only_template@<version>
,
if the project is:
- On GitLab.com
- Named
my-component
- In a personal namespace named
my-username
The templates directory and the suffix of the configuration file should be excluded from the referenced path.
If the project contains multiple components, then the file structure should be similar to:
├── README.md
├── .gitlab-ci.yml
└── templates/
└── all-scans.yml
└── secret-detection.yml
These components would be referenced with these paths:
gitlab.com/my-username/my-component/all-scans
gitlab.com/my-username/my-component/secret-detection
You can omit the filename in the path if the configuration file is named template.yml
.
For example, the following component could be referenced with gitlab.com/my-username/my-component/dast
:
├── README.md
├── .gitlab-ci.yml
├── templates/
│ └── dast
│ └── template.yml
Component configurations saved in any directory (deprecated)
Components configurations can be saved through the following directory structure, containing:
-
template.yml
: The component configuration, one file per component. If there is only one component, this file can be in the root of the project. If there are multiple components, each file must be in a separate subdirectory. -
README.md
: A documentation file explaining the details of all the components in the repository.
For example, if the project is on GitLab.com, named my-component
, and in a personal
namespace named my-username
:
-
Containing a single component and a simple pipeline to test the component, then the file structure might be:
├── template.yml ├── README.md └── .gitlab-ci.yml
The
.gitlab-ci.yml
file is not required for a CI/CD component to work, but testing the component in a pipeline in the project is recommended.This component is referenced with the path
gitlab.com/my-username/my-component@<version>
. -
Containing one default component and multiple sub-components, then the file structure might be:
├── template.yml ├── README.md ├── .gitlab-ci.yml ├── unit/ │ └── template.yml └── integration/ └── template.yml
These components are identified by these paths:
gitlab.com/my-username/my-component
gitlab.com/my-username/my-component/unit
gitlab.com/my-username/my-component/integration
It is possible to have a components repository with no default component, by having
no template.yml
in the root directory.
Additional notes:
Nesting of components is not possible. For example:
├── unit/
│ └── template.yml
│ └── another_folder/
│ └── nested_template.yml
Release a component
To create a release for a CI/CD component, use either:
- The
release
keyword in a CI/CD pipeline. Like in the component testing example, you can set a component to automatically be released after all tests pass in pipelines for new tags. - The UI for creating a release.
All released versions of the components are displayed in the CI/CD Catalog page for the given resource, providing users with information about official releases.
Components can be used without being released,
but only with a commit SHA or a branch name. To enable the use of tags or the ~latest
version keyword,
you must create a release.
Use a component in a CI/CD configuration
You can add a component to a CI/CD configuration with the include: component
keyword.
For example:
include:
- component: gitlab.example.com/my-namespace/my-component@1.0
inputs:
stage: build
The component is identified by a unique address in the form <fully-qualified-domain-name>/<component-path>@<specific-version>
,
where:
-
<fully-qualified-domain-name>
matches the GitLab host. You can only reference components in the same GitLab instance as your project. -
<component-path>
is the component project’s full path and directory where the component YAML file is located. -
<specific-version>
is the version of the component. In order of highest priority first, the version can be:- A branch name, for example
main
. - A commit SHA, for example
e3262fdd0914fa823210cdb79a8c421e2cef79d8
. - A tag, for example:
1.0
. If a tag and branch exist with the same name, the tag takes precedence over the branch. If a tag and commit SHA exist with the same name, the commit SHA takes precedence over the tag. -
~latest
, which is a special version that always points to the most recent released tag. Available only if the component has been released.
- A branch name, for example
For example, for a component repository located at gitlab-org/dast
on gitlab.com
,
the path:
-
gitlab.com/gitlab-org/dast@main
targets thetemplate.yml
in the root directory on themain
branch. -
gitlab.com/gitlab-org/dast@e3262fdd0914fa823210cdb79a8c421e2cef79d8
targets the same file for the specified commit SHA. -
gitlab.com/gitlab-org/dast@1.0
targets the same file for the1.0
tag. -
gitlab.com/gitlab-org/dast@~latest
targets the same file for the latest release. -
gitlab.com/gitlab-org/dast/api-scan@main
targets a different file, thetemplate.yml
in the/api-scan
directory in the component repository, for themain
branch.
Best practices
Avoid using global keywords
Avoid using global keywords in a component.
Using these keywords in a component affects all jobs in a pipeline, including jobs
directly defined in the main .gitlab-ci.yml
or in other included components.
As an alternative to global keywords, instead:
- Add the configuration directly to each job, even if it creates some duplication in the component configuration.
- Use the
extends
keyword in the component.
For example, using the default
keyword is not recommended:
# Not recommended
default:
image: ruby:3.0
rspec-1:
script: bundle exec rspec dir1/
rspec-2:
script: bundle exec rspec dir2/
Instead, you can:
-
Add the configuration to each job:
rspec-1: image: ruby:3.0 script: bundle exec rspec dir1/ rspec-2: image: ruby:3.0 script: bundle exec rspec dir2/
-
Use
extends
to reuse configuration:.rspec-image: image: ruby:3.0 rspec-1: extends: - .rspec-image script: bundle exec rspec dir1/ rspec-2: extends: - .rspec-image script: bundle exec rspec dir2/
Replace hard-coded values with inputs
Avoid hard-coding values in CI/CD components. Hard-coded values might force component users to need to review the component’s internal details and adapt their pipeline to work with the component.
A common keyword with problematic hard-coded values is stage
. If a component job’s
stage is set to a specific value, the pipeline using the component must define
the exact same stage. Additionally, if the component user wants to use a different stage,
they must override the configuration.
The preferred method is to use the input
keyword.
The component user can specify the exact value they need.
For example:
-
In the component configuration:
spec: inputs: stage: default: test --- unit-test: stage: $[[ inputs.stage ]] script: echo unit tests integration-test: stage: $[[ inputs.stage ]] script: echo integration tests
-
In the project using the component:
include: - component: gitlab.com/gitlab-org/ruby-test@1.0 inputs: stage: verify stages: [verify, deploy]
Replace custom CI/CD variables with inputs
When using CI/CD variables in a component, evaluate if the inputs
keyword
should be used instead. Avoid requiring a user to define custom variables to change a component’s
behavior. You should try to use inputs
for any component customization.
Inputs are explicitly defined in the component’s specs, and are better validated than variables. For example, if a required input is not passed to the component, GitLab returns a pipeline error. By contrast, if a variable is not defined, its value is empty, and there is no error.
For example, use inputs
instead of variables to let users change a scanner’s output format:
-
In the component configuration:
spec: inputs: scanner-output: default: json --- my-scanner: script: my-scan --output $[[ inputs.scanner-output ]]
-
In the project using the component:
include: - component: gitlab.example.com/my-scanner@1.0 inputs: scanner-output: yaml
In other cases, CI/CD variables are still preferred, including:
- Using predefined variables to automatically configure a component to match a user’s project.
- Requiring tokens or other sensitive values to be stored as masked or protected variables in project settings.
Use semantic versioning
When tagging and releasing new versions of components, you should use semantic versioning. Semantic versioning is the standard for communicating that a change is a major, minor, patch, or other kind of change.
You should use at least the major.minor
format, as this is widely understood. For example,
2.0
or 2.1
.
Other examples of semantic versioning:
1.0.0
2.1.3
1.0.0-alpha
3.0.0-rc1
Test the component
Testing CI/CD components as part of the development workflow is strongly recommended and helps ensure consistent behavior.
Test changes in a CI/CD pipeline (like any other project) by creating a .gitlab-ci.yml
in the root directory.
For example:
include:
# include the component located in the current project from the current SHA
- component: gitlab.com/$CI_PROJECT_PATH@$CI_COMMIT_SHA
inputs:
stage: build
stages: [build, test, release]
# Expect `component-job` is added.
# This example tests that the included component works as expected.
# You can inspect data generated by the component, use GitLab API endpoints or third-party tools.
ensure-job-added:
stage: test
image: badouralix/curl-jq
script:
- |
route="https://gitlab.com/api/v4/projects/$CI_PROJECT_ID/pipelines/$CI_PIPELINE_ID/jobs"
count=`curl --silent --header "PRIVATE-TOKEN: $API_TOKEN" $route | jq 'map(select(.name | contains("component-job"))) | length'`
if [ "$count" != "1" ]; then
exit 1
fi
# If we are tagging a release with a specific convention ("v" + number) and all
# previous checks succeeded, we proceed with creating a release automatically.
create-release:
stage: release
image: registry.gitlab.com/gitlab-org/release-cli:latest
rules:
- if: $CI_COMMIT_TAG =~ /^v\d+/
script: echo "Creating release $CI_COMMIT_TAG"
release:
tag_name: $CI_COMMIT_TAG
description: "Release $CI_COMMIT_TAG of components repository $CI_PROJECT_PATH"
After committing and pushing changes, the pipeline tests the component, then releases it if the test passes.
Convert a CI/CD template to a component
Any existing CI/CD template that you use in projects by using the include:
syntax
can be converted to a CI/CD component:
- Decide if you want the component to be part of an existing components repository to be grouped with other components, or create and set up a new components repository.
- Create a YAML file in the components repository according to the expected directory structure.
- Copy the content of the original template YAML file into the new component YAML file.
- Refactor the new component’s configuration to:
- Follow the best practices for components.
- Improve the configuration, for example by enabling merge request pipelines or making it more efficient.
- Leverage the
.gitlab-ci.yml
in the components repository to test changes to the component. - Tag and release the component.