[Mauiusers] Getting started

Andrus, Mr. Brian (Contractor) brian.andrus at nrlmry.navy.mil
Thu Sep 20 17:11:06 MDT 2007


All,

Ok, I have been using Torque and the default scheduler for a bit now and
I am finally getting around to trying out maui.
First, I prefer to use RPMs when I add/remove software from our system.
It is a bit cleaner. So I build Torque RPMs from the spec file that is
included with it.
For Maui, I came across a nice one at
http://hepunx.rl.ac.uk/~traylens/rpms/maui/
I decided to use the p17 src rpm and rebuild it. It seems to work very
well. I had to make a change in the maui.cfg file for my systems (use
the server name rather than localhost and update it to use RMCFG) and I
added the config path for my installation of Torque but that was it.

I installed the client package on the login node and the server+client
package on the server with Torque. So far so good. I turned off the
pbs_sched from Torque and started Maui. 
I am no able to run showq and see what is up. I submitted a job and it
went through. Cool!

But wait, the output of my job is simple, it says 'Hello, I am process X
of Y on node Z". My script requests 5 nodes with 1 processors per node,
but my output is only showing one node that it ran on!?
Hmm. I stop maui and restart pbs_sched and resubmit the job. It runs,
and the output is showing info from 5 nodes. So apparently there is more
that has to be done for Maui to work. 

Any ideas? My submission script follows:

-------------------------------------------------------
#!/bin/bash
#PBS -j oe
#PBS -l nodes=5:ppn=1
#PBS -N TestJob
#PBS -q medium
#PBS -o output.txt

mpiexec --bynode /data/andrus/hello
-------------------------------------------------------

cat output.txt (when using pbs_sched)
-------------------------------------------------------
Hello from process 0 of 5 on node n5
Hello from process 1 of 5 on node n4
Hello from process 2 of 5 on node n3
Hello from process 3 of 5 on node n2
Hello from process 4 of 5 on node n1
-------------------------------------------------------

cat output.txt (when using maui)
-------------------------------------------------------
Hello from process 0 of 1 on node n22
-------------------------------------------------------



Brian Andrus perotsystems 
Site Manager | Sr. Computer Scientist 
Naval Research Lab
7 Grace Hopper Ave, Monterey, CA  93943
Phone (831) 656-4839 | Fax (831) 656-4866 


-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.supercluster.org/pipermail/mauiusers/attachments/20070920/9219d90e/attachment.html


More information about the mauiusers mailing list