Yep, precisely.  If you send a bunch of these they should be run on
different nodes.
Also, if you tell it to use several processors it should run it on
several different processors, and it doesn't...


On 9/16/08, Abhishek Kulkarni <[EMAIL PROTECTED]> wrote:
>
>  On Tue, 2008-09-16 at 15:49 -0400, Daniel Gruner wrote:
>
> > Hi Hugh,
>  >
>  > I am still having some weird problems with moab/torque on my test xcpu
>  > cluster.  I mentioned some of these in a previous e-mail, but the
>  > query went unanswered, and since you wrote the script(s) perhaps you
>  > could help me debug this?  Here is the issue:
>  >
>  > I have 2 compute nodes, each with 2 cpus.  I submit several jobs to
>  > the queue using qsub:
>  >
>  > [EMAIL PROTECTED] xcpu]$ xstat
>  > n0000   tcp!10.10.0.10!6667     /Linux/x86_64   up      0
>  > n0001   tcp!10.10.0.11!6667     /Linux/x86_64   up      0
>  >
>  > [EMAIL PROTECTED] xcpu]$ showq
>  >
>  > active jobs------------------------
>  > JOBID                     USERNAME      STATE PROCS   REMAINING
>  >     STARTTIME
>  >
>  > 25.dgk3.chem.utoronto.ca     danny    Running     1    00:58:30  Mon
>  > Sep 15 10:11:07
>  > 26.dgk3.chem.utoronto.ca     danny    Running     1    00:58:30  Mon
>  > Sep 15 10:11:07
>  > 27.dgk3.chem.utoronto.ca     danny    Running     1    00:58:30  Mon
>  > Sep 15 10:11:07
>  > 28.dgk3.chem.utoronto.ca     danny    Running     1    00:58:30  Mon
>  > Sep 15 10:11:07
>  >
>  > 4 active jobs               4 of 4 processors in use by local jobs
>  > (100.00%)
>  >                             2 of 2 nodes active      (100.00%)
>  >
>  > eligible jobs----------------------
>  > JOBID              USERNAME      STATE PROCS     WCLIMIT
>  > QUEUETIME
>  >
>  >
>  > 0 eligible jobs
>  >
>  > blocked jobs-----------------------
>  > JOBID              USERNAME      STATE PROCS     WCLIMIT
>  > QUEUETIME
>  >
>  >
>  > 0 blocked jobs
>  >
>  > Total jobs:  4
>  >
>  > The job script is:
>  > #!/bin/bash
>  > #PBS -l nodes=1
>
>           ^^^^^^^
>  Isn't this supposed to mean, run "date" on 1 node?
>
>
>  > #XCPU -p
>  >
>  > date
>  >
>  >
>  > The weird thing is that all the jobs end up being executed on node
>  > n0000, as per the output:
>  >
>  > [EMAIL PROTECTED] xcpu]$ cat script.cmd.o25
>  > n0000: Mon Sep 15 10:11:42 UTC 2008
>  > [EMAIL PROTECTED] xcpu]$ cat script.cmd.o26
>  > n0000: Mon Sep 15 10:11:26 UTC 2008
>  > [EMAIL PROTECTED] xcpu]$ cat script.cmd.o27
>  > n0000: Mon Sep 15 10:13:03 UTC 2008
>  > [EMAIL PROTECTED] xcpu]$ cat script.cmd.o28
>  > n0000: Mon Sep 15 10:12:02 UTC 2008
>  >
>
> [snip]
>
>

Reply via email to