Workflow management in Condor

Workflows for e-Science: Scientific Workflows for Grids

Volumn , Issue , 2007, Pages 357-375

  Couvares, Peter ^a   Kosar, Tevfik ^b   Roy, Alain ^a   Weber, Jeff ^a   Wenger, Kent ^a  

^a University of Wisconsin Madison   (United States)

^b Louisiana State University   (United States)

Author keywords

[No Author keywords available]

Indexed keywords

EID: 84892356132     PISSN: None     EISSN: None     Source Type: Book    
DOI: 10.1007/978-1-84628-757-2_22     Document Type: Chapter

References (2)

1. Note that DAGMan assumes the batch system guarantees that it will not "lose" jobs after they have been successfully submitted. Currently, if the job is lost by the batch system after being successfully submitted by DAGMan, DAGMan will wait indefinitely for the status of the job in the queue to change. An explicit query for the status of submitted jobs (as opposed to waiting for the batch system to record job status changes) may be necessary to address this. Also, if a job languishes in the queue forever, DAGMan currently is not able to "timeout" and remove the job and mark it as failed. When removing jobs, detecting and responding to the failure of a remove operation (leaving a job "stuck" in the queue) is an interesting question.
2. In recent versions of Condor, the job can be edited to contain "noop job = true" which immediately terminates the job successfully.