Jobs might appear to continue running in Flow Manager when you attempt to kill the job (whether via the Flow Manager GUI or via the jkill command from a command prompt). When this happens, a message similar to the following appears in the jfd.log file:

Jan 17 14:41:45 16537 38 3 JFLSFExecutionAgent::checkReturnStatus: Failed to execute command , exited with <255>. Job <9999>: Job has already finished

When a job finishes, if you use the LSF command BHIST -L job-id, you will see two events that are similar to the following:

----------- 
Fri Mar  7 14:27:39: Done successfully. The CPU time used is 0.0 seconds; 
Fri Mar  7 14:27:40: Post job process done successfully; 
----------- 

The first message, Done successfully, means that the execution of the job itself is done. The second message, Post job process done successfully, means that the post-job clean up finished successfully as well.

The problem occurs because Platform Process Manager (the jfd executable file) does not receive the signal from the second LSF event, which indicates that the job is done. Because the application does not receive the signal, it does not consider the job complete. As such, the job still appears to be running.

To circumvent the issue, force Process Manager to look for the first event. Then edit the js.conf file and add the following line:

You must restart Process Manager (via the jfd executable file) for this new behavior to take effect. This change does not negatively affect the use of Process Manager or of any job flows.

Operating System and Release Information