Delete a Pipeline

You can delete a pipeline by running:

pachctl delete pipeline <pipeline_name>

To delete all of your pipelines (be careful with this feature), use the additional --all flag.

When you delete a pipeline:

  • Kubernetes deletes all resources associated with the pipeline - pods (if any), services, and replication controllers.
  • HPE Machine Learning Data Management deletes the user output repository with all its data as well as the system meta (stats) and spec (historical versions of the pipeline specification file) repositories.

You can use the --keep-repo flag to preserve the output repo with all its branches. However, important job metadata will still be deleted (including all historical versions of the pipeline specification file). As a result, you will not be able to recreate the deleted pipeline with the same name unless that repo is deleted first.


For example, if a pipeline “xyz” exists, then there is an output repo “xyz”. If a user deletes the pipeline with --keep-repo, the output repo “xyz” will remain, but the pipeline will be gone. If the user tries to create a new pipeline called “xyz”, it will fail (there is already an output repo with that name). For the pipeline creation to be successful, the user would have to delete repo “xyz” first.

You can use the output repo of a pipeline deleted with --keep-repo as an input repo and add more data.

When HPE Machine Learning Data Management cannot delete a pipeline with the standard command, you might need to enforce deletion using the --force flag. Because this option can break dependent components in your DAG, use this option withextreme caution.