On Tue, 9 Jun 2015 13:51:02 +0000 Dan Hyatt <[email protected]> wrote:
> I should explain the queues better. > > I have > > normal - default > HiMem -jobs needing up to 30GB per node (max 2 jobs per host) > Hi CPU - same except jobs run 100% cpu for weeks, so need 2 jobs per > host max > short - kills job after 60 min > vshort - kils job after 30 minutes > > so I > > qsub -q normal -cwd ........ > or > qsub -q HiMem -cwd ..... The above should work but is 'not the grid engine way'. Generally one sets grid engine up so that users request the resources they want (memory,cpu time, wall clock time) and grid engine routes them to an appropriate queue/host combo. > > and if normal is full it will go to HiMem, then try HiCPU, then short, > then very short... > I want it to wait until the queue is available that the job is assigned > to. I don't want it to try the next queue That looks like it should work. I'd try a qstat -j on the job to see if the hard queue_list reflects your request. -- William Hay <[email protected]>
pgpnmb7nNf8AU.pgp
Description: PGP signature
_______________________________________________ users mailing list [email protected] https://gridengine.org/mailman/listinfo/users
