- Enable pipelines for merged results
- Using Merge Trains
- Automatic pipeline cancellation
When you submit a merge request, you are requesting to merge changes from a source branch into a target branch. By default, the CI pipeline runs jobs against the source branch.
With pipelines for merged results, the pipeline runs as if the changes from the source branch have already been merged into the target branch.
If the pipeline fails due to a problem in the target branch, you can wait until the target is fixed and re-run the pipeline. This new pipeline will run as if the source is merged with the updated target, and you will not need to rebase.
The pipeline does not automatically run when the target branch changes. Only changes to the source branch trigger a new pipeline. If a long time has passed since the last successful pipeline, you may want to re-run it before merge, to ensure that the source changes can still be successfully merged into the target.
When the merge request can’t be merged, the pipeline runs against the source branch only. For example, when:
- The target branch has changes that conflict with the changes in the source branch.
- The merge request is a Draft merge request.
In these cases, the pipeline runs as a pipeline for merge requests
and is labeled as
detached. If these cases no longer exist, new pipelines will
again run against the merged results.
Any user who has developer permissions can run a pipeline for merged results.
To enable pipelines for merge results:
- You must have maintainer permissions.
- You must be using GitLab Runner 11.9 or later.
- You must not be using fast forward merges yet. To follow progress, see #58226.
To enable pipelines for merged results for your project:
- Configure your CI/CD configuration file so that the pipeline or individual jobs run for merge requests.
- Visit your project’s Settings > General and expand Merge requests.
- Check Enable merge trains and pipelines for merged results.
- Click Save changes.
Generally, this is a safer option than merging merge requests immediately, because your merge request will be evaluated with an expected post-merge result before the actual merge happens.
For more information, read the documentation on Merge Trains.
GitLab CI/CD can detect the presence of redundant pipelines, and will cancel them automatically in order to conserve CI resources.
When a user merges a merge request immediately within an ongoing merge train, the train will be reconstructed, as it will recreate the expected post-merge commit and pipeline. In this case, the merge train may already have pipelines running against the previous expected post-merge commit. These pipelines are considered redundant and will be automatically canceled.
Can be caused by some disabled feature flags. Please make sure that the following feature flags are enabled on your GitLab instance:
To check and set these feature flag values, please ask an administrator to:
Log into the Rails console of the GitLab instance:
sudo gitlab-rails console
Check if the flags are enabled or not:
If needed, enable the feature flags:
Since pipelines for merged results are a run on a merge ref of a merge request
refs/merge-requests/<iid>/merge), the Git reference could be overwritten at an
unexpected timing. For example, when a source or target branch is advanced.
In this case, the pipeline fails because of
fatal: reference is not a tree: error,
which indicates that the checkout-SHA is not found in the merge ref.
This behavior was improved at GitLab 12.4 by introducing Persistent pipeline refs. You should be able to create pipelines at any timings without concerning the error.