Thanks Reuti for your answer and pointing me in the right direction with your old mailing list message. The problem was that the number of slots per execution host was enforced with an RQS. If I set a value for slots as a consumable resource in each exechost definition, it works as expected.
For the record I also mention that I set 30 days as default duration and I use the first version of Open Grid Scheduler, which should be very close to Sun 6.2u5 I suppose. So the problem was a bad interaction between RQS and reservation. Thanks again Matthieu Melot -----Original Message----- From: Reuti [mailto:[email protected]] Sent: Tuesday, February 07, 2012 4:16 PM To: Melot Matthieu Cc: [email protected] Subject: Re: [gridengine users] Reservation and hard queue list Hi, Am 07.02.2012 um 21:25 schrieb Melot Matthieu: > I have a question on the ability to ask for a reservation that spans different queues. Here is the situation: > > Let's say that I have a queue defined on one node of 2 slots, I have a RQS that limit the number of slots per host to 2 > > I launch the following jobs (dummy.q has one execution host of 2 slots) > > qsub -N sleeper1 -pe mpi 1 -q dummy.q sleeper.sh 3600 > qsub -N sleeper2 -pe mpi 1 -q dummy.q sleeper.sh 3600 > qsub -N sleeper3 -pe mpi 1 -q dummy.q sleeper.sh 3600 NB: What is the default runtime which is assumed for the above jobs, as no h_rt is given? $ qconf -ssconf entry "default_duration". > I have jobs sleeper1 and sleeper2 running, > > and then I launch, > qsub -N reserv1 -pe mpi 2 -q dummy.q -R y -l h_rt=3600 sleeper.sh 3600 > > I have the intended behavior, namely that if I stop sleeper1, sleeper3 won't start, and if I stop sleeper2, reserv1 can start which is what I want. Also in $SGE_ROOT/$SGE_CELL/common/schedule, I can see that the keyword RESERVING appears, and confirm that all the stuff is working. > > Now, I delete everything and I relaunch sleeper1, 2 and 3 as before. > > Then I launch the job reserv2 that I want to run on a different queue (dummy2.q) that is again defined on the same host with two slots: > qsub -N reserv2 -pe mpi 2 -q dummy2.q -R y -l h_rt=3600 sleeper.sh 3600 > > In that situation I have no more effect of the reservation. If I stop sleeper1, sleeper3 will start immediately and prevent reserv2 to start. Also, in the schedule file, there are no more RESERVING keywords that appear. Which version of SGE are you using? I remember a similar thread: http://arc.liv.ac.uk/pipermail/gridengine-users/2010-July/031327.html -- Reuti > Do I make something wrong with the configuration? Is it because of the multiple instance queues / host, limited with an RQS that I don't have the reservation? If yes, what would be an alternative for the desired behavior? > > Thank you in advance for your help > > Regards > > Matthieu Melot > > > This message and any attachments are solely for the use of the intended recipients. They may contain privileged and/or confidential information or other information protected from disclosure. If you are not an intended recipient, you are hereby notified that you received this email in error and that any review, dissemination, distribution or copying of this email and any attachment is strictly prohibited. If you have received this email in error, please contact the sender and delete the message and any attachment from your system. > > > Thank You. > > _______________________________________________ > users mailing list > [email protected] > https://gridengine.org/mailman/listinfo/users _______________________________________________ users mailing list [email protected] https://gridengine.org/mailman/listinfo/users
