[GE users] reserving free ram?

cowlesav alan.cowles at duke.edu
Mon Nov 2 12:42:57 GMT 2009

Hey guys,

I am very new to SGE and management of a cluster environment, as I am 
usually as NAS admin. But I have a question that I haven't been able to 
get a definitive answer to on how to set and I thought I would throw it 
out to the community and see if someone had a way to set it up. I have 
currently a small 32 node blade cluster each with 16GB of ram per box. I 
have users that submit jobs that start out only requiring 3GB of ram but 
at points swell to 8GB of ram. Lets say they submit multiple copies of 
the same job. The qmaster will see a box totally free, submit the job, 
and at first consume 3GB of ram, the 2nd job goes in, and seeing the 
same box, with 13GB free, it submits it there, and then the 3rd job 
follows along, because there is still plenty of free ram for the 
starting jobs, all the way up to 5 jobs per node, with 1GB ram free. But 
when the first job starts to expand to that 8GB level, and it can't get 
there, it hangs, and all the other jobs are doomed as well.

I understand there is a way to set it, so that when you submit the job, 
it automagically reserves 8GB of ram even though the job is only using 
3, so that the qmaster will only assign 2 jobs to each node max.

Any help you guys can provide would be most appreciated.



To unsubscribe from this discussion, e-mail: [users-unsubscribe at gridengine.sunsource.net].

More information about the gridengine-users mailing list