Note: This discussion is about an older version of the COMSOL Multiphysics® software. The information provided may be out of date.

Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.

cluster computing mpi

Please login with a confirmed email address before reporting spam

Hello,

I am trying to run my models from comsol on our University's super computer, and having difficulty with the $PBS_NODEFILE command. Can the PBS_NODEFILE also be used to run generic MPI programs?

Thank you,
Kelley

5 Replies Last Post 29.08.2013, 22:43 GMT-4
Niklas Rom COMSOL Employee

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 14.02.2013, 04:02 GMT-5
Hi, see slide 19 in the attached slideshow.

-------------------------------------------
cds.comsol.com/mg/9511ca78cae71f.zip
Estimated size: 0.7 MB
This link expires February 21, 2013. Please make sure to download before that date.

Included files:
- cluster_install_linux_43a.pptx
-------------------------------------------
Niklas


Hello,

I am trying to run my models from comsol on our University's super computer, and having difficulty with the $PBS_NODEFILE command. Can the PBS_NODEFILE also be used to run generic MPI programs?

Thank you,
Kelley


Hi, see slide 19 in the attached slideshow. ------------------------------------------- http://cds.comsol.com/mg/9511ca78cae71f.zip Estimated size: 0.7 MB This link expires February 21, 2013. Please make sure to download before that date. Included files: - cluster_install_linux_43a.pptx ------------------------------------------- Niklas [QUOTE] Hello, I am trying to run my models from comsol on our University's super computer, and having difficulty with the $PBS_NODEFILE command. Can the PBS_NODEFILE also be used to run generic MPI programs? Thank you, Kelley [/QUOTE]

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 14.02.2013, 14:26 GMT-5
Thank you for redirecting me to that power point. I have seen that before, but I had forgotten about it.
I am still having trouble running anything more complicated than

comsol batch -inputfile in.mph -outputfile out.mph

With version 4.3 I was able to use clustersimple command successfully but not any longer. I keep getting the same error " $PBS_NODEFILE: ambiguous redirect"

This is how I have it in my script

total=$(wc -l < $PBS_NODEFILE)
uniq -c $PBS_NODEFILE

comsol -nn 4 -clustersimple batch -inputfile 2.14_simplemodel.mph -outputfile 2.14_out.mph -batchlog Parallel.log

Do I need to include < -f $PBS_NODEFILE > in my final command line as well?

Thank you
Thank you for redirecting me to that power point. I have seen that before, but I had forgotten about it. I am still having trouble running anything more complicated than comsol batch -inputfile in.mph -outputfile out.mph With version 4.3 I was able to use clustersimple command successfully but not any longer. I keep getting the same error " $PBS_NODEFILE: ambiguous redirect" This is how I have it in my script total=$(wc -l < $PBS_NODEFILE) uniq -c $PBS_NODEFILE comsol -nn 4 -clustersimple batch -inputfile 2.14_simplemodel.mph -outputfile 2.14_out.mph -batchlog Parallel.log Do I need to include < -f $PBS_NODEFILE > in my final command line as well? Thank you

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 15.02.2013, 13:45 GMT-5
Is anyone familiar with running comsol on a cluster from command line? I could use some help!
Thank you!

comsol models will not run on more than 1 node. The MPD commands do not work, and it seems like the problem is communication with the nodes.
Does MPD need to be checked in the GUI before I can run a model through command line in a cluster?
Are there new commands for COMSOL 4.3a for the MPD communication and host file?
Our computer science department is not familiar with Comsol, so any help would be appreciated.

The commands I am using below

qsub -q generic -I -l nodes=4:ppn12
module load comsol/43a
total=$(w -l < $PBS_NODEFILE)
uniq -c $PBS_NODEFILE
comsol -nn 4 mpd boot -f $PBS_NODEFILE -mpirsh ssh --verbose
comsol -nn 4 -np 12 batch -inputfile in.mph -outputfile out.mph

Thank you so much,
Kelley
Is anyone familiar with running comsol on a cluster from command line? I could use some help! Thank you! comsol models will not run on more than 1 node. The MPD commands do not work, and it seems like the problem is communication with the nodes. Does MPD need to be checked in the GUI before I can run a model through command line in a cluster? Are there new commands for COMSOL 4.3a for the MPD communication and host file? Our computer science department is not familiar with Comsol, so any help would be appreciated. The commands I am using below qsub -q generic -I -l nodes=4:ppn12 module load comsol/43a total=$(w -l < $PBS_NODEFILE) uniq -c $PBS_NODEFILE comsol -nn 4 mpd boot -f $PBS_NODEFILE -mpirsh ssh --verbose comsol -nn 4 -np 12 batch -inputfile in.mph -outputfile out.mph Thank you so much, Kelley

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 18.02.2013, 11:58 GMT-5
I've got a response from the Comsol Support, so I thought I would post it here in case anyone else has a similar questions

The problem here is likely related to the new Hydra launching method introduced in COMSOL 4.3a. Normally Hydra should automatically detect the necessary settings from the cluster environment, and with this method it is not necessary to launch mpd:s manually (as was necessary in 4.3 and earlier versions). When using PBS, it should be possible (with the Hydra launcher) to replace the two comsol commands ("comsol mpd boot" and "comsol batch") with this single comsol batch command:

comsol -nn 4 -np 12 batch -inputfile in.mph -outputfile out.mph -mpiarg -rmk -mpiarg pbs -clustersimple

Alternatively, if the Hydra method does not work satisfactorily, it is possible to revert back to the mpd launch method used in COMSOL 4.3. To use the old method, the only thing needed is to add the "-mpd" option to the comsol batch command. The COMSOL commands in your PBS script would now be:

comsol -nn 4 mpd boot -f $PBS_NODEFILE -mpirsh ssh --verbose
comsol -mpd -nn 4 -np 12 batch -inputfile in.mph -outputfile out.mp
comsol mpd allexit
I've got a response from the Comsol Support, so I thought I would post it here in case anyone else has a similar questions The problem here is likely related to the new Hydra launching method introduced in COMSOL 4.3a. Normally Hydra should automatically detect the necessary settings from the cluster environment, and with this method it is not necessary to launch mpd:s manually (as was necessary in 4.3 and earlier versions). When using PBS, it should be possible (with the Hydra launcher) to replace the two comsol commands ("comsol mpd boot" and "comsol batch") with this single comsol batch command: comsol -nn 4 -np 12 batch -inputfile in.mph -outputfile out.mph -mpiarg -rmk -mpiarg pbs -clustersimple Alternatively, if the Hydra method does not work satisfactorily, it is possible to revert back to the mpd launch method used in COMSOL 4.3. To use the old method, the only thing needed is to add the "-mpd" option to the comsol batch command. The COMSOL commands in your PBS script would now be: comsol -nn 4 mpd boot -f $PBS_NODEFILE -mpirsh ssh --verbose comsol -mpd -nn 4 -np 12 batch -inputfile in.mph -outputfile out.mp comsol mpd allexit

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 29.08.2013, 22:43 GMT-4
thank your for sharing
thank your for sharing

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.