Troubleshooting CI/CD (FREE)

GitLab provides several tools to help make troubleshooting your pipelines easier.

This guide also lists common issues and possible solutions.

Verify syntax

An early source of problems can be incorrect syntax. The pipeline shows a yaml invalid badge and does not start running if any syntax or formatting problems are found.

Edit gitlab-ci.yml with the pipeline editor

The pipeline editor is the recommended editing experience (rather than the single file editor or the Web IDE). It includes:

  • Code completion suggestions that ensure you are only using accepted keywords.
  • Automatic syntax highlighting and validation.
  • The CI/CD configuration visualization, a graphical representation of your .gitlab-ci.yml file.

If you prefer to use another editor, you can use a schema like the Schemastore gitlab-ci schema with your editor of choice.

Verify syntax with CI Lint tool

The CI Lint tool is a simple way to ensure the syntax of a CI/CD configuration file is correct. Paste in full .gitlab-ci.yml files or individual jobs configuration, to verify the basic syntax.

When a .gitlab-ci.yml file is present in a project, you can also use the CI Lint tool to simulate the creation of a full pipeline. It does deeper verification of the configuration syntax.

Verify variables

A key part of troubleshooting CI/CD is to verify which variables are present in a pipeline, and what their values are. A lot of pipeline configuration is dependent on variables, and verifying them is one of the fastest ways to find the source of a problem.

Export the full list of variables available in each problematic job. Check if the variables you expect are present, and check if their values are what you expect.

GitLab CI/CD documentation

The complete .gitlab-ci.yml reference contains a full list of every keyword you can use to configure your pipelines.

You can also look at a large number of pipeline configuration examples and templates.

Documentation for pipeline types

Some pipeline types have their own detailed usage guides that you should read if you are using that type:

Troubleshooting Guides for CI/CD features

Troubleshooting guides are available for some CI/CD features and related topics:

Common CI/CD issues

A lot of common pipeline issues can be fixed by analyzing the behavior of the rules or only/except configuration. You shouldn't use these two configurations in the same pipeline, as they behave differently. It's hard to predict how a pipeline runs with this mixed behavior.

If your rules or only/except configuration makes use of predefined variables like CI_PIPELINE_SOURCE, CI_MERGE_REQUEST_ID, you should verify them as the first troubleshooting step.

Jobs or pipelines don't run when expected

The rules or only/except keywords are what determine whether or not a job is added to a pipeline. If a pipeline runs, but a job is not added to the pipeline, it's usually due to rules or only/except configuration issues.

If a pipeline does not seem to run at all, with no error message, it may also be due to rules or only/except configuration, or the workflow: rules keyword.

If you are converting from only/except to the rules keyword, you should check the rules configuration details carefully. The behavior of only/except and rules is different and can cause unexpected behavior when migrating between the two.

The common if clauses for rules can be very helpful for examples of how to write rules that behave the way you expect.

Two pipelines run at the same time

Two pipelines can run when pushing a commit to a branch that has an open merge request associated with it. Usually one pipeline is a merge request pipeline, and the other is a branch pipeline.

This situation is usually caused by the rules configuration, and there are several ways to prevent duplicate pipelines.

A job is not in the pipeline

GitLab determines if a job is added to a pipeline based on the only/except or rules defined for the job. If it didn't run, it's probably not evaluating as you expect.

No pipeline or the wrong type of pipeline runs

Before a pipeline can run, GitLab evaluates all the jobs in the configuration and tries to add them to all available pipeline types. A pipeline does not run if no jobs are added to it at the end of the evaluation.

If a pipeline did not run, it's likely that all the jobs had rules or only/except that blocked them from being added to the pipeline.

If the wrong pipeline type ran, then the rules or only/except configuration should be checked to make sure the jobs are added to the correct pipeline type. For example, if a merge request pipeline did not run, the jobs may have been added to a branch pipeline instead.

It's also possible that your workflow: rules configuration blocked the pipeline, or allowed the wrong pipeline type.

A job runs unexpectedly

A common reason a job is added to a pipeline unexpectedly is because the changes keyword always evaluates to true in certain cases. For example, changes is always true in certain pipeline types, including scheduled pipelines and pipelines for tags.

The changes keyword is used in combination with only/except or rules). It's recommended to use changes with rules or only/except configuration that ensures the job is only added to branch pipelines or merge request pipelines.

"fatal: reference is not a tree" error

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 default 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 occurs because the previous pipeline cannot find a checkout-SHA (which is associated with the pipeline record) from the example branch that the commit history has already been overwritten by the force-push. Similarly, Merged results pipelines 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. The runner fetches the persistent pipeline ref and gets source code from the checkout-SHA.
  4. When the pipeline finishes, its persistent ref is cleaned up in a background process.

Merge request pipeline messages

The merge request pipeline widget shows information about the pipeline status in a merge request. It's displayed above the ability to merge status widget.

"Checking ability to merge automatically" message

There is a known issue where a merge request can be stuck with the Checking ability to merge automatically message.

If your merge request has this message and it does not disappear after a few minutes, you can try one of these workarounds:

  • Refresh the merge request page.
  • Close & Re-open the merge request.
  • Rebase the merge request with the /rebase quick action.
  • If you have already confirmed the merge request is ready to be merged, you can merge it with the /merge quick action.

"Checking pipeline status" message

This message is shown when the merge request has no pipeline associated with the latest commit yet. This might be because:

  • GitLab hasn't finished creating the pipeline yet.
  • You are using an external CI service and GitLab hasn't heard back from the service yet.
  • You are not using CI/CD pipelines in your project.
  • You are using CI/CD pipelines in your project, but your configuration prevented a pipeline from running on the source branch for your merge request.
  • The latest pipeline was deleted (this is a known issue).
  • The source branch of the merge request is on a private fork.

After the pipeline is created, the message updates with the pipeline status.

Merge request status messages

The merge request status widget shows the Merge button and whether or not a merge request is ready to merge. If the merge request can't be merged, the reason for this is displayed.

If the pipeline is still running, the Merge button is replaced with the Merge when pipeline succeeds button.

If Merge Trains are enabled, the button is either Add to merge train or Add to merge train when pipeline succeeds. (PREMIUM)

"A CI/CD pipeline must run and be successful before merge" message

This message is shown if the Pipelines must succeed setting is enabled in the project and a pipeline has not yet run successfully. This also applies if the pipeline has not been created yet, or if you are waiting for an external CI service. If you don't use pipelines for your project, then you should disable Pipelines must succeed so you can accept merge requests.

"Merge blocked: pipeline must succeed. Push a new commit that fixes the failure" message

This message is shown if the merge request pipeline, merged results pipeline, or merge train pipeline has failed or been canceled.

If a merge request pipeline or merged result pipeline was canceled or failed, you can:

  • Re-run the entire pipeline by clicking Run pipeline in the pipeline tab in the merge request.
  • Retry only the jobs that failed. If you re-run the entire pipeline, this is not necessary.
  • Push a new commit to fix the failure.

If the merge train pipeline has failed, you can:

  • Check the failure and determine if you can use the /merge quick action to immediately add the merge request to the train again.
  • Re-run the entire pipeline by clicking Run pipeline in the pipeline tab in the merge request, then add the merge request to the train again.
  • Push a commit to fix the failure, then add the merge request to the train again.

If the merge train pipeline was canceled before the merge request was merged, without a failure, you can:

  • Add it to the train again.

Project group/project not found or access denied

This message is shown if configuration is added with include and one of the following:

  • The configuration refers to a project that can't be found.
  • The user that is running the pipeline is unable to access any included projects.

To resolve this, check that:

  • The path of the project is in the format my-group/my-project and does not include any folders in the repository.
  • The user running the pipeline is a member of the projects that contain the included files. Users must also have the permission to run CI/CD jobs in the same projects.

"The parsed YAML is too big" message

This message displays when the YAML configuration is too large or nested too deeply. YAML files with a large number of includes, and thousands of lines overall, are more likely to hit this memory limit. For example, a YAML file that is 200kb is likely to hit the default memory limit.

To reduce the configuration size, you can:

  • Check the length of the expanded CI/CD configuration in the pipeline editor's merged YAML tab. Look for duplicated configuration that can be removed or simplified.
  • Move long or repeated script sections into standalone scripts in the project.
  • Use parent and child pipelines to move some work to jobs in an independent child pipeline.

On a self-managed instance, you can increase the size limits.

Pipeline warnings

Pipeline configuration warnings are shown when you:

"Job may allow multiple pipelines to run for a single action" warning

When you use rules with a when clause without an if clause, multiple pipelines may run. Usually this occurs when you push a commit to a branch that has an open merge request associated with it.

To prevent duplicate pipelines, use workflow: rules or rewrite your rules to control which pipelines can run.

Console workaround if job using resource_group gets stuck (FREE SELF)

# find resource group by name
resource_group = Project.find_by_full_path('...').resource_groups.find_by(key: 'the-group-name')
busy_resources = resource_group.resources.where('build_id IS NOT NULL')

# identify which builds are occupying the resource
# (I think it should be 1 as of today)

# it's good to check why this build is holding the resource.
# Is it stuck? Has it been forcefully dropped by the system?
# free up busy resources
busy_resources.update_all(build_id: nil)

How to get help

If you are unable to resolve pipeline issues, you can get help from: