[GE users] SGE- head node goes into swap on job submission with too much memory-requirements:
srflaxu40 at gmail.com
Wed Oct 27 22:15:23 BST 2010
[ The following text is in the "iso-8859-1" character set. ]
[ Your display is set for the "ISO-8859-10" character set. ]
[ Some special characters may be displayed incorrectly. ]
I'm using an SGE grid and I constantly come across the problem of exceeding the head node's memory and causing the head node to go into swap. Although you should submit to the cluster, there is no 100% guaranteed way to ensure this never happens. Therefore, is there a program or application out there that can prevent this from occurring- a sort of second-check to ensure no job submitted on the head node exceeds the memory and causes the node to go into swap? If it does have a chance or does potentially exceed current memory usage, it's terminated immediately. This way it never goes into swap and your cluster is up and running. Thanks.
More information about the gridengine-users