• Variables usage
  • .gitlab-ci.yml file
  • config.toml file
• Expansion mechanisms
  • GitLab internal variable expansion mechanism
    • Nested variable expansion
      • Enabling the nested variable expansion feature
  • GitLab Runner internal variable expansion mechanism
  • Execution shell environment
• Persisted variables
• Variables with an environment scope

Where variables can be used

As it’s described in the CI/CD variables docs, you can define many different variables. Some of them can be used for all GitLab CI/CD features, but some of them are more or less limited.

This document describes where and how the different types of variables can be used.

Variables usage

There are two places defined variables can be used. On the:

1. GitLab side, in .gitlab-ci.yml.
2. The GitLab Runner side, in config.toml.

.gitlab-ci.yml file

config.toml file

You can read more about config.toml in the GitLab Runner docs.

Expansion mechanisms

There are three expansion mechanisms:

• GitLab
• GitLab Runner
• Execution shell environment

GitLab internal variable expansion mechanism

The expanded part needs to be in a form of $variable, or ${variable} or %variable%. Each form is handled in the same way, no matter which OS/shell handles the job, because the expansion is done in GitLab before any runner gets the job.

Nested variable expansion

GitLab expands job variable values recursively before sending them to the runner. For example:

- BUILD_ROOT_DIR: '${CI_BUILDS_DIR}'
- OUT_PATH: '${BUILD_ROOT_DIR}/out'
- PACKAGE_PATH: '${OUT_PATH}/pkg'

If nested variable expansion is:

• Disabled: the runner receives ${BUILD_ROOT_DIR}/out/pkg. This is not a valid path.
• Enabled: the runner receives a valid, fully-formed path. For example, if ${CI_BUILDS_DIR} is /output, then PACKAGE_PATH would be /output/out/pkg.

References to unavailable variables are left intact. In this case, the runner attempts to expand the variable value at runtime. For example, a variable like CI_BUILDS_DIR is known by the runner only at runtime.

Enabling the nested variable expansion feature

This feature comes with the :variable_inside_variable feature flag disabled by default.

To enable this feature, ask a GitLab administrator with Rails console access to run the following command:

# For the instance
Feature.enable(:variable_inside_variable)
# For a single project
Feature.enable(:variable_inside_variable, Project.find(<project id>))

To disable it:

# For the instance
Feature.disable(:variable_inside_variable)
# For a single project
Feature.disable(:variable_inside_variable, Project.find(<project id>))

GitLab Runner internal variable expansion mechanism

• Supported: project/group variables, .gitlab-ci.yml variables, config.toml variables, and variables from triggers, pipeline schedules, and manual pipelines.
• Not supported: variables defined inside of scripts (e.g., export MY_VARIABLE="test").

The runner uses Go’s os.Expand() method for variable expansion. It means that it handles only variables defined as $variable and ${variable}. What’s also important, is that the expansion is done only once, so nested variables may or may not work, depending on the ordering of variables definitions, and whether nested variable expansion is enabled in GitLab.

Execution shell environment

This is an expansion phase that takes place during the script execution. Its behavior depends on the shell used (bash, sh, cmd, PowerShell). For example, if the job’s script contains a line echo $MY_VARIABLE-${MY_VARIABLE_2}, it should be properly handled by bash/sh (leaving empty strings or some values depending whether the variables were defined or not), but don’t work with Windows’ cmd or PowerShell, since these shells use a different variables syntax.

Supported:

• The script may use all available variables that are default for the shell (e.g., $PATH which should be present in all bash/sh shells) and all variables defined by GitLab CI/CD (project/group variables, .gitlab-ci.yml variables, config.toml variables, and variables from triggers and pipeline schedules).
• The script may also use all variables defined in the lines before. So, for example, if you define a variable export MY_VARIABLE="test":
  • In before_script, it works in the subsequent lines of before_script and all lines of the related script.
  • In script, it works in the subsequent lines of script.
  • In after_script, it works in subsequent lines of after_script.

In the case of after_script scripts, they can:

• Only use variables defined before the script within the same after_script section.
• Not use variables defined in before_script and script.

These restrictions exist because after_script scripts are executed in a separated shell context.

Persisted variables

The following variables are known as “persisted”:

• CI_PIPELINE_ID
• CI_JOB_ID
• CI_JOB_TOKEN
• CI_JOB_STARTED_AT
• CI_BUILD_ID
• CI_BUILD_TOKEN
• CI_REGISTRY_USER
• CI_REGISTRY_PASSWORD
• CI_REPOSITORY_URL
• CI_DEPLOY_USER
• CI_DEPLOY_PASSWORD

They are:

• Supported for definitions where the “Expansion place” is:
  • Runner.
  • Script execution shell.
• Not supported:
  • For definitions where the “Expansion place” is GitLab.
  • In the only and except variables expressions.

Some of the persisted variables contain tokens and cannot be used by some definitions due to security reasons.

Variables with an environment scope

Variables defined with an environment scope are supported. Given that there is a variable $STAGING_SECRET defined in a scope of review/staging/*, the following job that is using dynamic environments is created, based on the matching variable expression:

my-job:
  stage: staging
  environment:
    name: review/$CI_JOB_STAGE/deploy
  script:
    - 'deploy staging'
  rules:
    - if: $STAGING_SECRET == 'something'