StarCluster - Mailing List Archive

Specifying resources on SGE

From: Amirhossein Kiani <no email>
Date: Mon, 5 Dec 2011 12:21:38 -0800

Dear all,

I imagine this question might be more of a SGE question, but thought you guys might have dealt with this issue.

I am running jobs with high CPU and memory usage and the jobs start with low resource consumption and grow as time goes on.
I have been specifying resource limitation such as mem_free but the issue I run into is that the grid engine looks at the state of each node when jobs are being submitted and sees that there is enough memory for the job on the nodes. Therefore it ends up submitting multiple jobs per node. The jobs run into memory issues when their memory size grows and get killed due to memory consumption. Additionally, every job has many threads and I would like to specify the number of CPUs that need to be only allocated to each job.

Do you guys happen to know a way around this?

Many thanks!
Received on Mon Dec 05 2011 - 15:21:39 EST
This archive was generated by hypermail 2.3.0.


Sort all by: