TURBOMOLE Users Forum

TURBOMOLE Modules => Aoforce and Numforce => Topic started by: vvallet on March 11, 2008, 10:12:38 PM

Title: Problem with parallel NumForce in turbomole 5.10
Post by: vvallet on March 11, 2008, 10:12:38 PM

Dear colleagues,

I am trying to run turbomole in parallel with 4 processors by requesting 4 processors on 1 node through the PBS queuing system:
qsub -l nodes=1:ppn=4

The program stops with the following message:

 running on compute-0-0.local
 date: Tue Mar 11 21:57:27 CET 2008

execute parallel run using NODEFILE: /opt/torque/aux//1729.cluster1.univ-lille1.fr
There are 4 free nodes:  compute-0-0.local-1 compute-0-0.local-2 compute-0-0.local-3 compute-0-0.local-4
all nodes will be used for calculation

Error logging in on compute-0-0.local-1

It runs on 4 processors located on node compute-0-0.local but tries to connect to compute-0-0.local-1 to ompute-0-0.local-4, although these names do not exist. It looks to me as if there is a problem with the way the node list is built?

Thanks in advance for your help.

Valérie Vallet

Title: Re: Problem with parallel NumForce in turbomole 5.10
Post by: antti_karttunen on March 12, 2008, 07:40:02 AM

Hello,

Considering the node naming scheme, you seem to use Rocks clusters (or something similar). Perhaps your problem is related to the issue reported here: http://www.turbo-forum.com/index.php?topic=140.msg362#msg362
If this is the case, I suppose TURBOMOLE crew can provide a fixed NumForce script.

Title: Re: Problem with parallel NumForce in turbomole 5.10
Post by: vvallet on March 12, 2008, 09:58:31 AM

Hello,

Thanks a lot for pointing out your previous post. I followed your suggestion and changed all "-" separators to "_" in the NumForce script.

Regards,

Valérie