How to rerun samples

Introduction

It is recommended that you have read and completed the basic pipestat tutorial as well as configuring pipestat backends.

This document will briefly cover details about how to configure pipestat in looper:

Learning objectives

How do I rerun samples?

There may be a time when you need to re-submit samples. Looper's rerun command can be used to achieve this. By default, the rerun command will gather any samples with a failed status and submit again.

However, Looper does not create flags for samples. It only reads them. Creating flags is the job of the pipeline via either the pipeline manager (e.g. pypiper) or pipestat.

Therefore, to create submission flags, you must:

Configure looper to use pipestat

Add the flag_file_dir key and a flag directory path under the pipestat key within the looper config:

.looper.yaml

pep_config: project_config_pipestat.yaml
output_dir: results
pipeline_interfaces:
  - ./pipeline_interface1_sample_pipestat.yaml
pipestat:
  flag_file_dir: output/results_pipeline
  results_file_path: tmp_pipestat_results.yaml

Ensure your pipeline is using pipestat to set the status upon sample processing.

currentPipestatManager.set_status(record_identifier="sample1", status_identifier="running")
currentPipestatManager.set_status(record_identifier="sample2", status_identifier="failed")

More info on setting status can be found here.

Now that your pipeline is setting statuses, you can resubmit failed jobs with:

looper rerun

By default, this command will only submit failed jobs. However, you can use the --ignore-flags flag to resubmit all jobs.

looper rerun --ignore-flags