Cleaning Up After E155x Jobs

Back to E155x Home Page
The E155x analysis job has a nasty habit of leaving behind Shared Memory (SM) segments if it is killed or otherwise terminates abnormally. These left over SMs continue to use up real memory, leaving less available to other processes. Performance degrades, and jobs eventually fail because they could not obtain the necessary memory. There have been many cases where batch machines are so low on memory that they are effectively dead. The system administrator could not log in to reboot, and had to power cycle the affected machines, something that people will be reluctant to do especially at night and on weekends.

It is therefore important to clean up after each job, and not allow the effects to accumulate. The procedure is simple. You must do it yourself, because others cannot remove your SMs.

Extra cleanup commands do not do any harm, so go ahead and issue it in case you have doubts. The command to look for SM should be obvious if you look inside the e155cleanup script.

A scan is made of all the interactive and batch machines, i.e. morgans and pintos, every morning. Anyone with a SM is sent an E-mail notice. This is a reminder for those who might occasionally miss one SM. Everyone should proactively clean up without waiting for the reminder; the notice may come too late to avoid powering cycling a computer.

Lee Sorrell