The following (and tarball) is from a workflow with most tasks removed due to data reuse. The cleanup algorithm is getting confused because a file listed as transfer=True in a job exists on a remote site, but not the output site. A staging job is added to transfer from the remote site to the output site, but then a cleanup job is being added.

2014.02.20 16:20:08.605 PST: [DEBUG] Clustering 2 cleanup nodes at level 2 with cluster size 1
2014.02.20 16:20:08.605 PST: [DEBUG] Creating a clustered cleanup job named clean_up_staging_level_2_0 consisting of 1 nodes
2014.02.20 16:20:08.605 PST: [DEBUG] Creating a clustered cleanup job named clean_up_staging_level_2_1 consisting of 1 nodes
2014.02.20 16:20:08.605 PST: [DEBUG] For cleanup job clean_up_staging_level_2_0 the associated compute job is create_dir_pipeline_0_staging
2014.02.20 16:20:08.606 PST: [FATAL ERROR] java.lang.RuntimeException: Unable to determine cleanup url for lfn HN001_addrepl.bam at site null
at edu.isi.pegasus.planner.refiner.cleanup.Cleanup.createCleanupJob(Cleanup.java:213)
at edu.isi.pegasus.planner.refiner.cleanup.InPlace.addCleanUpJobs(InPlace.java:532)
at edu.isi.pegasus.planner.refiner.cleanup.InPlace.addCleanupJobs(InPlace.java:298)
at edu.isi.pegasus.planner.refiner.CleanupEngine.addCleanupJobs(CleanupEngine.java:69)
at edu.isi.pegasus.planner.refiner.MainEngine.runPlanner(MainEngine.java:279)
at edu.isi.pegasus.planner.client.CPlanner.executeCommand(CPlanner.java:646)
at edu.isi.pegasus.planner.client.CPlanner.executeCommand(CPlanner.java:366)
at edu.isi.pegasus.planner.client.CPlanner.main(CPlanner.java:254)