Is there any good docs, or anyone's personal notes to help me figure out configuring the queues. Looking at qmon, there are many settings (many more command line) but I am not clear which to use when. I have read a lot of good detailed docs out there, that have helped me in other areas. But I am not clear regarding the settings for the queues.

I meant to expand on the prior question:
I have different needs,
a high mem queue, which is set up to run on 16 server.
A high CPU queue that needs to run one job per server or per socket. (jobs that use a CPU 100% for days or weeks). Trying to figure out which setting limit the number of jobs per socket or server.


a normal queue that is default,

a short queue: for fast jobs, under 1 hour. A higher priority. what is the difference between a soft limit and a hard limit. Hard limit is set to 60 minutes. I believe this kills the job after 60 min, but what does the soft limit do?

Am I correct in understanding that priority....jobs start at negative priorities and increase as they are not scheduled... the highest a user can set is priority 0, but I should set jobs I want to wait to -60, and jobs I want to run quickly at -10?

thanks in advance


On 7/11/2014 9:09 AM, Dan Hyatt wrote:
Hello,

I have a newly installed grid engine, with about 1200 slots (cores).
I have two users with jobs right now, one user has 30 jobs active and about 500 queued up in pending.
Yet my grid engine is 99% idle.

The user states she is not specifying a queue. But it is sending all jobs to my "high cpu" queue that only has 30 slots. My normal queue has 1200 slots.

how do I change the "default" queue if they do not specify "qsub -q normal ..."

Thanks



--
Dan Hyatt
Division of Statistical Genomics
Washington University School of Medicine
4444 Forest Park Blvd, Campus Box 8506
St. Louis, MO  63108
314 747 4767 (o)
314 473 8713 (c)
[email protected]

_______________________________________________
users mailing list
[email protected]
https://gridengine.org/mailman/listinfo/users

Reply via email to