[rock-user] que jobs die out problem, Please help

Dear ROCKS?SGE masters,

Recently, I'm begin to submit some big jobs with qsub command. Each job 
suppose to run for one week or so. But many of them die out after only 1-2 
days. And some nodes become un-avaliable to the front end..
My question is where is the SGE log files, or system log files I can check 
and trouble shooting?
What kind of problem it could be? From the Ganglia windows, I can see that 
the swap memory is high arround the time having problem. ( 2G memory, the 
swap usage was ~600M when it has problem) Is it might be a memory issue? If 
I could break down the jobs, does it help?

Thanks and regards,

Leo