I have not used the program myself, but from a brief reading of the documentation at clusterresources, I believe you might be seeing a "feature" of pbsdsh. It looks like that program works on a per node basis and not on a per process basis.
A beter test would be to have your script simply sleep for 10 seconds or something and queue up as many copies as you have processors and see if you get one running on each processor. You could also test it using an mpi program, as those tend to work more like you think they should. On 3/14/06, Olof Mattsson <[EMAIL PROTECTED]> wrote: > Hi! > > We're two students from University of Skövde Sweden that studies network > and systemadministration and right now are we working on a project to > build and evaluate a Linux computing cluster. Our first approach is > using OSCAR and it's working quite good right now. At the moment we've > only got five computing nodes and one master node but later this week > we're gonna add eight more nodes. > The master node is a dual AMD 2400+ MP, 2GB RAM, with two nic, eth1 is > connected to our lab network and eth0 is connected to a Summit4. The > first node is exactly the same but with only one nic. The other four > nodes are AMD 1900+, 1GB RAM. Everything is connected through the > Summit. We run Fedora Core 3 and OSCAR 4.2. > > Works great but when we submit jobs we can only run it on one CPU per > node or only on the dual CPU node but on both CPUs. > > We've tried these different commands to submit our job (the script > submit_pi below, nothing fancy but we're only testing so far): > > qsub -l nodes=5,ncpus=1,walltime=10:00 submit_pi > Works but runs only on one CPU per node > > qsub -l nodes=5,ncpus=2,walltime=10:00 submit_pi > Runs the job only on the dual CPU node but on both cpus > > qsub -l nodes=1:ppn=2+4:ppn=1,walltime=10:00 submit_pi > Runs the job on the four singel CPU nodes > > #!/bin/bash > #submit_pi > > #PBS -N test_job > #PBS -o test_job_stdout.txt > #PBS -e test_job_stderr.txt > #PBS -q workq > > time pbsdsh /home/oscartst/pibench/pi_c50 > > #Nothing more to see here > > Tried a couple more but with same result or the job doesn't start at all. > It's not crucial if we can't get it to work since we've only got one > dual CPU computing node but it would be great if any of you could help us to > get it to work. > > Regards > Olof Mattsson > > > > > > ------------------------------------------------------- > This SF.Net email is sponsored by xPML, a groundbreaking scripting language > that extends applications into web and mobile media. Attend the live webcast > and join the prime developer group breaking into this new coding territory! > http://sel.as-us.falkag.net/sel?cmdlnk&kid0944&bid$1720&dat1642 > _______________________________________________ > Oscar-users mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/oscar-users > ------------------------------------------------------- This SF.Net email is sponsored by xPML, a groundbreaking scripting language that extends applications into web and mobile media. Attend the live webcast and join the prime developer group breaking into this new coding territory! http://sel.as-us.falkag.net/sel?cmd=lnk&kid0944&bid$1720&dat1642 _______________________________________________ Oscar-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/oscar-users
