Still trying to see why my jobs are being pre-empted on machine A and re-started on machine B, I thought maybe they are dying themselves due to a run error. However, the output and error log files are empty due to buffering, even when the job is killed via condor_rm (and there should be a lot of text in those log files). Is there a way to force periodic flushing of the log files, and force a flush when a job is killed by the user?
Ralph Finch
Calif. Dept. of Water Resources
![[Computer Systems Lab]](https://www-auth.cs.wisc.edu/pics/csl_logo.gif){kind=logo}