Re: submitting a job in a queue

From: Gengbin Zheng (gzheng_at_ks.uiuc.edu)
Date: Thu Jun 03 2004 - 11:33:25 CDT

Hi Giovanni,

You can write a little script to generate the "nodelist" file on the fly,
such as the following:

#!/bin/tcsh

echo "group main ++shell ssh > nodelist

set list=`cat $PBS_NODEFILE`

foreach h ( $list )
 echo "host $h" >> nodelist
end

Gengbin

On Wed, 2 Jun 2004, Giovanni wrote:

> Hi all,
> I would like to start a parallel simulation with namd using charmrun on
> an linux-alpha cluster.
> The software is working perfectly for local jobs but if I want to use
> remote machines also,
> the only thing I can do is to submit my job to a queue and wait for free
> nodes.
> Then, I don't know in advance on which machines my simulation will run.
> I guess, first, that all the names of the machines in the cluster should
> be in the "nodelist" file, but is this enough to get it working ?
>
> I am a bit suspicious because, when I run NAMD on our smaller cluster
> (intel-linux and no queues) I have to copy the directory containing the
> starting files to an identical path in all remote nodes that are in my
> nodelist.
> Also, the executable namd2 must be in the same directory on all machines
> that I decide to use
> Any suggestion ?
> Thanks
>
> G.
>
> --
> ========================================
> Giovanni Bellesia
> Theory and Computation Group
> Department of Chemistry
> University College Dublin
> Belfield, Dublin 4, Ireland
> ========================================
>
>

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:37:39 CST