[torqueusers] Help! In Mac OS X All Jobs Are Being Sent to OneNode

Fabrizio Salvatore p.salvatore at rhul.ac.uk
Mon Mar 14 02:10:12 MST 2005


Hi Steven,

again, my experience is with a Linux pbs server, but what I expect it should
happen in your case as well is that in the directory where your pbs server
is installed (in my case is /var/spool/pbs/server_priv, I'm  not sure where
it is in yours) you need to have a file called 'nodes' where you specify the
setup of your batch nodes. The minimun is to write there the node name and
the number of cpus, in the following format

node01 np=2
node02 np=2
....
nodeXX np=2

Then restart your server using the option -t create to create a new
serverdb. At this point in qmgr, if you print the status of any given node,
you sould see something like this:

Qmgr: p n nodeXX
#
# Create and define node nodeXX
#
set node nodexx state=free
.....
set node nodexx status arch=MacOS (or something similar, if this is a MacOS
queue)
.....
set node nodexx status +=ncpus=2
.....

Hope this helps,

Cheers,

				Fab

-----Original Message-----
From: torqueusers-bounces at supercluster.org
[mailto:torqueusers-bounces at supercluster.org] On Behalf Of Steven Saunders
Sent: 14 March 2005 01:21
To: torqueusers at supercluster.org
Subject: RE: [torqueusers] Help! In Mac OS X All Jobs Are Being Sent to
OneNode



--- Fabrizio Salvatore <p.salvatore at rhul.ac.uk> wrote:
> Ous cluster is a Linux one and not a Mac OS one, but what you say 
> suggests me that you do not have the option set to distribute the jobs 
> evenly. in qmgr:
> 
> > qmgr
> Qmgr: p s
> 
> do you see:
> set server resources_default.neednodes = 1
> set server resources_default.nodect = 1
> 
> If not, set these parameters and try again.

Thanks for your response. Those settings helped, jobs are now being sent to
all the nodes except now only one job is being sent to each node. 

Each of my execution nodes has 2 cpus, and each of my jobs only requires
only one cpu. So I need each node to get two jobs at a time. However, when I
submit many jobs, only one job is getting sent to each node, and the other
jobs are waiting in the queue and are not sent to a node until the single
job already on the node is finished.

I've tried using 

        qsub -l nodes=1:ppn=1 myjob

but that doesn't make any difference.

How can I get it to give each node as many jobs as it has CPUs?

Thanks.



Find local movie times and trailers on Yahoo! Movies.
http://au.movies.yahoo.com _______________________________________________
torqueusers mailing list
torqueusers at supercluster.org
http://supercluster.org/mailman/listinfo/torqueusers




More information about the torqueusers mailing list