Compliance pipelines

DETAILS: Tier: Ultimate Offering: SaaS, self-managed

Introduced in GitLab 13.9, disabled behind ff_evaluate_group_level_compliance_pipeline feature flag.

Enabled by default in GitLab 13.11.

Feature flag removed in GitLab 14.2.

Group owners can configure a compliance pipeline in a project separate to other projects. By default, the compliance pipeline configuration (for example, .compliance-gitlab-ci.yml) is run instead of the pipeline configuration (for example, .gitlab-ci.yml) of labeled projects.

However, the compliance pipeline configuration can reference the .gitlab-ci.yml file of the labeled projects so that:

The compliance pipeline can also run jobs of labeled project pipelines. This allows for centralized control of pipeline configuration.
Jobs and variables defined in the compliance pipeline can't be changed by variables in the labeled project's .gitlab-ci.yml file.

NOTE: Because of a known issue, project pipelines must be included first at the top of compliance pipeline configuration to prevent projects overriding settings downstream.

For more information, see:

Example configuration for help configuring a compliance pipeline that runs jobs from labeled project pipeline configuration.
The Create a compliance pipeline tutorial.

Effect on labeled projects

Users have no way of knowing that a compliance pipeline has been configured and might be confused why their own pipelines are not running at all, or include jobs that they did not define themselves.

When authoring pipelines on a labeled project, there is no indication that a compliance pipeline has been configured. The only marker at the project level is the compliance framework label itself, but the label does not say whether the framework has a compliance pipeline configured or not.

Therefore, communicate with project users about compliance pipeline configuration to reduce uncertainty and confusion.

Configure a compliance pipeline

To configure a compliance pipeline:

On the left sidebar, select Search or go to and find your group.
Select Settings > General.
Expand the Compliance frameworks section.
In Compliance pipeline configuration (optional), add the path to the compliance framework configuration. Use the path/file.y[a]ml@group-name/project-name format. For example:
- .compliance-ci.yml@gitlab-org/gitlab.
- .compliance-ci.yaml@gitlab-org/gitlab.

This configuration is inherited by projects where the compliance framework label is applied. In projects with the applied compliance framework label, the compliance pipeline configuration is run instead of the labeled project's own pipeline configuration.

The user running the pipeline in the labeled project must at least have the Reporter role on the compliance project.

When used to enforce scan execution, this feature has some overlap with scan execution policies. We have not unified the user experience for these two features. For details on the similarities and differences between these features, see Enforce scan execution.

Example configuration

The following example .compliance-gitlab-ci.yml includes the include keyword to ensure labeled project pipeline configuration is also executed.

include:  # Execute individual project's configuration (if project contains .gitlab-ci.yml)
  - project: '$CI_PROJECT_PATH'
    file: '$CI_CONFIG_PATH'
    ref: '$CI_COMMIT_SHA' # Must be defined or MR pipelines always use the use default branch
    rules:
      - if: $CI_PROJECT_PATH != "my-group/project-1" # Must be the hardcoded path to the project that hosts this configuration.

# Allows compliance team to control the ordering and interweaving of stages/jobs.
# Stages without jobs defined will remain hidden.
stages:
  - pre-compliance
  - build
  - test
  - pre-deploy-compliance
  - deploy
  - post-compliance

variables:  # Can be overridden by setting a job-specific variable in project's local .gitlab-ci.yml
  FOO: sast

sast:  # None of these attributes can be overridden by a project's local .gitlab-ci.yml
  variables:
    FOO: sast
  image: ruby:2.6
  stage: pre-compliance
  rules:
    - if: $CI_COMMIT_BRANCH && $CI_OPEN_MERGE_REQUESTS && $CI_PIPELINE_SOURCE == "push"
      when: never
    - when: always  # or when: on_success
  allow_failure: false
  before_script:
    - "# No before scripts."
  script:
    - echo "running $FOO"
  after_script:
    - "# No after scripts."

sanity check:
  image: ruby:2.6
  stage: pre-deploy-compliance
  rules:
    - if: $CI_COMMIT_BRANCH && $CI_OPEN_MERGE_REQUESTS && $CI_PIPELINE_SOURCE == "push"
      when: never
    - when: always  # or when: on_success
  allow_failure: false
  before_script:
    - "# No before scripts."
  script:
    - echo "running $FOO"
  after_script:
    - "# No after scripts."

audit trail:
  image: ruby:2.7
  stage: post-compliance
  rules:
    - if: $CI_COMMIT_BRANCH && $CI_OPEN_MERGE_REQUESTS && $CI_PIPELINE_SOURCE == "push"
      when: never
    - when: always  # or when: on_success
  allow_failure: false
  before_script:
    - "# No before scripts."
  script:
    - echo "running $FOO"
  after_script:
    - "# No after scripts."

The rules configuration in the include definition avoids circular inclusion in case the compliance pipeline must be able to run in the host project itself. You can leave it out if your compliance pipeline only ever runs in labeled projects.

Compliance pipelines and custom pipeline configuration hosted externally

The example above assumes that all projects host their pipeline configuration in the same project. If any projects use configuration hosted externally to the project:

The include section in the example compliance pipeline configuration must be adjusted. For example, using include:rules:

include:
  # If the custom path variables are defined, include the project's external config file.
  - project: '$PROTECTED_PIPELINE_CI_PROJECT_PATH'
    file: '$PROTECTED_PIPELINE_CI_CONFIG_PATH'
    ref: '$PROTECTED_PIPELINE_CI_REF'
    rules:
      - if: $PROTECTED_PIPELINE_CI_PROJECT_PATH && $PROTECTED_PIPELINE_CI_CONFIG_PATH && $PROTECTED_PIPELINE_CI_REF
  # If any custom path variable is not defined, include the project's internal config file as normal.
  - project: '$CI_PROJECT_PATH'
    file: '$CI_CONFIG_PATH'
    ref: '$CI_COMMIT_SHA'
    rules:
      - if: $PROTECTED_PIPELINE_CI_PROJECT_PATH == null || $PROTECTED_PIPELINE_CI_CONFIG_PATH == null || $PROTECTED_PIPELINE_CI_REF == null

CI/CD variables must be added to projects with external pipeline configuration. In this example:
- PROTECTED_PIPELINE_CI_PROJECT_PATH: The path to the project hosting the configuration file, for example group/subgroup/project.
- PROTECTED_PIPELINE_CI_CONFIG_PATH: The path to the configuration file in the project, for example path/to/.gitlab-ci.yml.
- PROTECTED_PIPELINE_CI_REF: The ref to use when retrieving the configuration file, for example main.

Compliance pipelines in merge requests originating in project forks

When a merge request originates in a fork, the branch to be merged usually only exists in the fork. When creating such a merge request against a project with compliance pipelines, the above snippet fails with a Project <project-name> reference <branch-name> does not exist! error message. This error occurs because in the context of the target project, $CI_COMMIT_REF_NAME evaluates to a non-existing branch name.

To get the correct context, use $CI_MERGE_REQUEST_SOURCE_PROJECT_PATH instead of $CI_PROJECT_PATH. This variable is only available in merge request pipelines.

For example, for a configuration that supports both merge request pipelines originating in project forks and branch pipelines, you need to combine both include directives with rules:if:

include:  # Execute individual project's configuration (if project contains .gitlab-ci.yml)
  - project: '$CI_MERGE_REQUEST_SOURCE_PROJECT_PATH'
    file: '$CI_CONFIG_PATH'
    ref: '$CI_COMMIT_REF_NAME'
    rules:
      - if: $CI_PIPELINE_SOURCE == 'merge_request_event'
  - project: '$CI_PROJECT_PATH'
    file: '$CI_CONFIG_PATH'
    ref: '$CI_COMMIT_REF_NAME'
    rules:
      - if: $CI_PIPELINE_SOURCE != 'merge_request_event'

Ensure compliance jobs are always run

Compliance pipelines use GitLab CI/CD to give you an incredible amount of flexibility for defining any sort of compliance jobs you like. Depending on your goals, these jobs can be configured to be:

Modified by users.
Non-modifiable.

Generally, if a value in a compliance job:

Is set, it cannot be changed or overridden by project-level configurations.
Is not set, a project-level configuration may be set.

Either might be wanted or not depending on your use case.

The following are a few best practices for ensuring that these jobs are always run exactly as you define them and that downstream, project-level pipeline configurations cannot change them:

Add a rules:when:always block to each of your compliance jobs. This ensures they are non-modifiable and are always run.
Explicitly set any variables the job references. This:
- Ensures that project-level pipeline configurations do not set them and alter their behavior. For example, see before_script and after_script configuration in the example configuration.
- Includes any jobs that drive the logic of your job.
Explicitly set the container image to run the job in. This ensures that your script steps execute in the correct environment.
Explicitly set any relevant GitLab pre-defined job keywords. This ensures that your job uses the settings you intend and that they are not overridden by project-level pipelines.

Avoid parent and child pipelines in GitLab 14.7 and earlier

NOTE: This advice does not apply to GitLab 14.8 and later because a fix added compatibility for combining compliance pipelines, and parent and child pipelines.

Compliance pipelines start on the run of every pipeline in a labeled project. This means that if a pipeline in the labeled project triggers a child pipeline, the compliance pipeline runs first. This can trigger the parent pipeline, instead of the child pipeline.

Therefore, in projects with compliance frameworks, you should replace parent-child pipelines with the following:

Direct include statements that provide the parent pipeline with child pipeline configuration.
Child pipelines placed in another project that are run using the trigger API rather than the parent-child pipeline feature.

This alternative ensures the compliance pipeline does not re-start the parent pipeline.

Troubleshooting

Compliance jobs are overwritten by target repository

If you use the extends statement in a compliance pipeline configuration, compliance jobs are overwritten by the target repository job. For example, you could have the following .compliance-gitlab-ci.yml configuration:

"compliance job":
  extends:
    - .compliance_template
  stage: build

.compliance_template:
  script:
    - echo "take compliance action"

You could also have the following .gitlab-ci.yml configuration:

"compliance job":
  stage: test
  script:
    - echo "overwriting compliance action"

This configuration results in the target repository pipeline overwriting the compliance pipeline, and you get the following message: overwriting compliance action.

To avoid overwriting a compliance job, don't use the extends keyword in compliance pipeline configuration. For example, you could have the following .compliance-gitlab-ci.yml configuration:

"compliance job":
  stage: build
  script:
    - echo "take compliance action"

You could also have the following .gitlab-ci.yml configuration:

"compliance job":
  stage: test
  script:
    - echo "overwriting compliance action"

This configuration doesn't overwrite the compliance pipeline and you get the following message: take compliance action.

Prefilled variables are not shown

Because of a known issue, compliance pipelines in GitLab 15.3 and later can prevent prefilled variables from appearing when manually starting a pipeline.

To workaround this issue, use ref: '$CI_COMMIT_SHA' instead of ref: '$CI_COMMIT_REF_NAME' in the include: statement that executes the individual project's configuration.

The example configuration has been updated with this change:

include:
  - project: '$CI_PROJECT_PATH'
    file: '$CI_CONFIG_PATH'
    ref: '$CI_COMMIT_SHA'