CI/CD pipelines

Introduced in GitLab 8.8.

NOTE: Tip: Watch the "Mastering continuous software development" webcast to see a comprehensive demo of a GitLab CI/CD pipeline.

Pipelines are the top-level component of continuous integration, delivery, and deployment.

Pipelines comprise:

  • Jobs, which define what to do. For example, jobs that compile or test code.
  • Stages, which define when to run the jobs. For example, stages that run tests after stages that compile the code.

Jobs are executed by Runners. Multiple jobs in the same stage are executed in parallel, if there are enough concurrent runners.

If all jobs in a stage succeed, the pipeline moves on to the next stage.

If any job in a stage fails, the next stage is not (usually) executed and the pipeline ends early.

In general, pipelines are executed automatically and require no intervention once created. However, there are also times when you can manually interact with a pipeline.

A typical pipeline might consist of four stages, executed in the following order:

  • A build stage, with a job called compile.
  • A test stage, with two jobs called test1 and test2.
  • A staging stage, with a job called deploy-to-stage.
  • A production stage, with a job called deploy-to-prod.

NOTE: Note: If you have a mirrored repository that GitLab pulls from, you may need to enable pipeline triggering in your project's Settings > Repository > Pull from a remote repository > Trigger pipelines for mirror updates.

Types of pipelines

Pipelines can be configured in many different ways:

Configure a pipeline

Pipelines and their component jobs and stages are defined in the CI/CD pipeline configuration file for each project.

For a list of configuration options in the CI pipeline file, see the GitLab CI/CD Pipeline Configuration Reference.

You can also configure specific aspects of your pipelines through the GitLab UI. For example:

View pipelines

You can find the current and historical pipeline runs under your project's CI/CD > Pipelines page. You can also access pipelines for a merge request by navigating to its Pipelines tab.

Pipelines index page

Clicking a pipeline will bring you to the Pipeline Details page and show the jobs that were run for that pipeline. From here you can cancel a running pipeline, retry jobs on a failed pipeline, or delete a pipeline.

Starting in GitLab 12.3, a link to the latest pipeline for the last commit of a given branch is available at /project/pipelines/[branch]/latest. Also, /project/pipelines/latest will redirect you to the latest pipeline for the last commit on the project's default branch.

Starting in GitLab 13.0, you can filter the pipeline list by:

  • Trigger author
  • Branch name

Run a pipeline manually

Pipelines can be manually executed, with predefined or manually-specified variables.

You might do this if the results of a pipeline (for example, a code build) are required outside the normal operation of the pipeline.

To execute a pipeline manually:

  1. Navigate to your project's CI/CD > Pipelines.
  2. Click on the Run Pipeline button.
  3. On the Run Pipeline page:
    1. Select the branch to run the pipeline for in the Create for field.
    2. Enter any environment variables required for the pipeline run.
    3. Click the Create pipeline button.

The pipeline will execute the jobs as configured.

Run a pipeline by using a URL query string

Introduced in GitLab 12.5.

You can use a query string to pre-populate the Run Pipeline page. For example, the query string .../pipelines/new?ref=my_branch&var[foo]=bar&file_var[file_foo]=file_bar will pre-populate the Run Pipeline page with:

  • Run for field: my_branch.
  • Variables section:
    • Variable:
      • Key: foo
      • Value: bar
    • File:
      • Key: file_foo
      • Value: file_bar

The format of the pipelines/new URL is:


The following parameters are supported:

  • ref: specify the branch to populate the Run for field with.
  • var: specify a Variable variable.
  • file_var: specify a File variable.

For each var or file_var, a key and value are required.

Add manual interaction to your pipeline

Introduced in GitLab 8.15.

Manual actions, configured using the when:manual parameter, allow you to require manual interaction before moving forward in the pipeline.

You can do this straight from the pipeline graph. Just click the play button to execute that particular job.

For example, your pipeline might start automatically, but it requires manual action to deploy to production. In the example below, the production stage has a job with a manual action.

Pipelines example

Start multiple manual actions in a stage

Introduced in GitLab 11.11.

Multiple manual actions in a single stage can be started at the same time using the "Play all manual" button. Once the user clicks this button, each individual manual action will be triggered and refreshed to an updated status.

This functionality is only available:

  • For users with at least Developer access.
  • If the stage contains manual actions.

Delete a pipeline

Introduced in GitLab 12.7.

Users with owner permissions in a project can delete a pipeline by clicking on the pipeline in the CI/CD > Pipelines to get to the Pipeline Details page, then using the Delete button.

Pipeline Delete Button

CAUTION: Warning: Deleting a pipeline will expire all pipeline caches, and delete all related objects, such as builds, logs, artifacts, and triggers. This action cannot be undone.

Pipeline quotas

Each user has a personal pipeline quota that tracks the usage of shared runners in all personal projects. Each group has a usage quota that tracks the usage of shared runners for all projects created within the group.

When a pipeline is triggered, regardless of who triggered it, the pipeline quota for the project owner's namespace is used. In this case, the namespace can be the user or group that owns the project.

How pipeline duration is calculated

Total running time for a given pipeline excludes retries and pending (queued) time.

Each job is represented as a Period, which consists of:

  • Period#first (when the job started).
  • Period#last (when the job finished).

A simple example is:

  • A (1, 3)
  • B (2, 4)
  • C (6, 7)

In the example:

  • A begins at 1 and ends at 3.
  • B begins at 2 and ends at 4.
  • C begins at 6 and ends at 7.

Visually, it can be viewed as:

0  1  2  3  4  5  6  7

The union of A, B, and C is (1, 4) and (6, 7). Therefore, the total running time is:

(4 - 1) + (7 - 6) => 4

Pipeline security on protected branches

A strict security model is enforced when pipelines are executed on protected branches.

The following actions are allowed on protected branches only if the user is allowed to merge or push on that specific branch:

  • Run manual pipelines (using the Web UI or pipelines API).
  • Run scheduled pipelines.
  • Run pipelines using triggers.
  • Trigger manual actions on existing pipelines.
  • Retry or cancel existing jobs (using the Web UI or pipelines API).

Variables marked as protected are accessible only to jobs that run on protected branches, preventing untrusted users getting unintended access to sensitive information like deployment credentials and tokens.

Runners marked as protected can run jobs only on protected branches, preventing untrusted code from executing on the protected runner and preserving deployment keys and other credentials from being unintentionally accessed. In order to ensure that jobs intended to be executed on protected runners will not use regular runners, they must be tagged accordingly.

View jobs in a pipeline

When you access a pipeline, you can see the related jobs for that pipeline.

Clicking an individual job will show you its job log, and allow you to:

  • Cancel the job.
  • Retry the job.
  • Erase the job log.

See why a job failed

Introduced in GitLab 10.7.

When a pipeline fails or is allowed to fail, there are several places where you can find the reason:

  • In the pipeline graph, on the pipeline detail view.
  • In the pipeline widgets, in the merge requests and commit pages.
  • In the job views, in the global and detailed views of a job.

In each place, if you hover over the failed job you can see the reason it failed.

Pipeline detail

In GitLab 10.8 and later, you can also see the reason it failed on the Job detail page.

The order of jobs in a pipeline

The order of jobs in a pipeline depends on the type of pipeline graph.

The order of severity is:

  • failed
  • warning
  • pending
  • running
  • manual
  • scheduled
  • canceled
  • success
  • skipped
  • created

For example:

Pipeline mini graph sorting

Group jobs in a pipeline

Introduced in GitLab 8.12.

If you have many similar jobs, your pipeline graph becomes long and hard to read.

You can automatically group similar jobs together. If the job names are formatted in a certain way, they will be collapsed into a single group in regular pipeline graphs (not the mini graphs).

You'll know when a pipeline has grouped jobs if you don't see the retry or cancel button inside them. Hovering over them will show the number of grouped jobs. Click to expand them.

Grouped pipelines

To create a group of jobs, in the CI/CD pipeline configuration file, separate each job name with a number and one of the following:

  • A slash (/), for example, test 1/3, test 2/3, test 3/3.
  • A colon (:), for example, test 1:3, test 2:3, test 3:3.
  • A space, for example test 0 3, test 1 3, test 2 3.

You can use these symbols interchangeably.

For example, these three jobs will be in a group named build ruby:

build ruby 1/3:
  stage: build
  - echo "ruby1"

build ruby 2/3:
  stage: build
  - echo "ruby2"

build ruby 3/3:
  stage: build
  - echo "ruby3"

In the pipeline, the result is a group named build ruby with three jobs:

Job group

The jobs will be ordered by comparing the numbers from left to right. You usually want the first number to be the index and the second number to be the total.

This regular expression evaluates the job names: \d+[\s:\/\\]+\d+\s*.

Specifying variables when running manual jobs

Introduced in GitLab 12.2.

When running manual jobs you can supply additional job specific variables.

You can do this from the job page of the manual job you want to run with additional variables. To access this page, click on the name of the manual job in the pipeline view, not the play ({play}) button.

This is useful when you want to alter the execution of a job that uses custom environment variables. Adding a variable name (key) and value here will override the value defined in the UI or .gitlab-ci.yml, for a single run of the manual job.

Manual job variables

Delay a job

Introduced in GitLab 11.4.

When you do not want to run a job immediately, you can use the when:delayed parameter to delay a job's execution for a certain period.

This is especially useful for timed incremental rollout where new code is rolled out gradually.

For example, if you start rolling out new code and:

  • Users do not experience trouble, GitLab can automatically complete the deployment from 0% to 100%.
  • Users experience trouble with the new code, you can stop the timed incremental rollout by canceling the pipeline and rolling back to the last stable version.

Pipelines example

Expand and collapse job log sections

Introduced in GitLab 12.0.

Job logs are divided into sections that can be collapsed or expanded. Each section will display the duration.

In the following example:

  • Two sections are collapsed and can be expanded.
  • Three sections are expanded and can be collapsed.

Collapsible sections

Custom collapsible sections

You can create collapsible sections in job logs by manually outputting special codes that GitLab will use to determine what sections to collapse:

  • Section start marker: section_start:UNIX_TIMESTAMP:SECTION_NAME\r\e[0K + TEXT_OF_SECTION_HEADER
  • Section end marker: section_end:UNIX_TIMESTAMP:SECTION_NAME\r\e[0K

You must add these codes to the script section of the CI configuration. For example, using echo:

    - echo -e "section_start:`date +%s`:my_first_section\r\e[0KHeader of the 1st collapsible section"
    - echo 'this line should be hidden when collapsed'
    - echo -e "section_end:`date +%s`:my_first_section\r\e[0K"

In the example above:

  • date +%s: The Unix timestamp (for example 1560896352).
  • my_first_section: The name given to the section.
  • \r\e[0K: Prevents the section markers from displaying in the rendered (colored) job log, but they are displayed in the raw job log. To see them, in the top right of the job log, click {doc-text} (Show complete raw).
    • \r: carriage return.
    • \e[0K: clear line ANSI escape code.

Sample raw job log:

section_start:1560896352:my_first_section\r\e[0KHeader of the 1st collapsible section
this line should be hidden when collapsed

Visualize pipelines

Introduced in GitLab 8.11.

Pipelines can be complex structures with many sequential and parallel jobs.

To make it easier to understand the flow of a pipeline, GitLab has pipeline graphs for viewing pipelines and their statuses.

Pipeline graphs can be displayed in two different ways, depending on the page you access the graph from.

NOTE: Note: GitLab capitalizes the stages' names in the pipeline graphs.

Regular pipeline graphs

Regular pipeline graphs show the names of the jobs in each stage. Regular pipeline graphs can be found when you are on a single pipeline page. For example:

Pipelines example

Multi-project pipeline graphs help you visualize the entire pipeline, including all cross-project inter-dependencies. (PREMIUM)

Pipeline mini graphs

Pipeline mini graphs take less space and can tell you at a quick glance if all jobs passed or something failed. The pipeline mini graph can be found when you navigate to:

  • The pipelines index page.
  • A single commit page.
  • A merge request page.

Pipeline mini graphs allow you to see all related jobs for a single commit and the net result of each stage of your pipeline. This allows you to quickly see what failed and fix it.

Stages in pipeline mini graphs are collapsible. Hover your mouse over them and click to expand their jobs.

Mini graph Mini graph expanded
Pipelines mini graph Pipelines mini graph extended

Pipeline success and duration charts

  • Introduced in GitLab 3.1.1 as Commit Stats, and later renamed to Pipeline Charts.
  • Renamed to CI / CD Analytics in GitLab 12.8.

GitLab tracks the history of your pipeline successes and failures, as well as how long each pipeline ran. To view this information, go to Analytics > CI / CD Analytics.

View successful pipelines:

Successful pipelines

View pipeline duration history:

Pipeline duration

Pipeline badges

Pipeline status and test coverage report badges are available and configurable for each project. For information on adding pipeline badges to projects, see Pipeline badges.

Pipelines API

GitLab provides API endpoints to:

Troubleshooting fatal: reference is not a tree:

Introduced in GitLab 12.4.

Previously, you'd have encountered unexpected pipeline failures when you force-pushed a branch to its remote repository. To illustrate the problem, suppose you've had the current workflow:

  1. A user creates a feature branch named example and pushes it to a remote repository.
  2. A new pipeline starts running on the example branch.
  3. A user rebases the example branch on the latest master branch and force-pushes it to its remote repository.
  4. A new pipeline starts running on the example branch again, however, the previous pipeline (2) fails because of fatal: reference is not a tree: error.

This is because the previous pipeline cannot find a checkout-SHA (which associated with the pipeline record) from the example branch that the commit history has already been overwritten by the force-push. Similarly, Pipelines for merged results might have failed intermittently due to the same reason.

As of GitLab 12.4, we've improved this behavior by persisting pipeline refs exclusively. To illustrate its life cycle:

  1. A pipeline is created on a feature branch named example.
  2. A persistent pipeline ref is created at refs/pipelines/<pipeline-id>, which retains the checkout-SHA of the associated pipeline record. This persistent ref stays intact during the pipeline execution, even if the commit history of the example branch has been overwritten by force-push.
  3. GitLab Runner fetches the persistent pipeline ref and gets source code from the checkout-SHA.
  4. When the pipeline finished, its persistent ref is cleaned up in a background process.