In a cluster running jobs of varied length, a user running only short jobs may always have a higher priority than a user running a long job. This can happen when historical run time decay is applied, decreasing the impact of the completed short jobs but not the longer job that is still running. To correct this, you can configure LSF to decay the run time of a job that is still running in the same manner historical run time decays.
Once a job is complete, the decayed run time is transferred to the historical run time where the decay continues. This equalizes the effect of short and long running jobs on user dynamic priority.
Running badmin reconfig or restarting mbatchd during a job’s run time results in the decayed run time being recalculated. When a suspended job using run time decay is resumed, the decay time is based on the elapsed time.