starting cfx solver 19.2 on hpc

Svetlana · April 29, 2019, 09:11

When i start cfx5solve 19.2 on a cluster, it partitions the mesh and things, comes to '+--------------+ Solver +--------------+' and then stops there for many hours doing nothing. I am not at all sure how to troubleshoot it. Can it be a problem with the licence?

Gert-Jan · April 29, 2019, 09:50

Stops for hours? Do you mean that it is doing something but that you don't know what it is doing, and then after several hours continues normally? E.g. have you included radiation? I have seen cases where it took hours for CFX to do the raytracing, but then continued normally.......

Or is it really doing nothing and is waiting for you to kill the job? Does CFX leave a text file with an error in the directory that is named after the output file?

evcelica · April 29, 2019, 10:43

This could be the result of many things. I do not think it is the licence though, that would give you an error. Usually something is not right with the configuration of distributed CFX.

You should run small test model distributed until you get everything set up correctly.

Uninstall old versions or set the environmental variables correctly. Ensure "hosts" file is correct. Allow program and processes through firewall including MPI processes. perhaps start on only 2 machines distributed, as there may be a problem with only 1 of the nodes. This could be difficult to troubleshoot.

Svetlana · April 29, 2019, 17:56

No radiation. It does not continue normally. Iteration number 1 does not begin.

I'll test with smaller distributed runs and write back. Thanks!

kjetil · May 8, 2019, 10:22

It is possibly a problem with IntelMPI. You might need to install the "old" version, which is found in the installation directory.

Svetlana · May 8, 2019, 22:02

You're right, I was loading a conflicting version of openmpi in my bashrc because I needed this for another script. Now I can see how this was a confusing and poor place to put that line in. I removed it and everything started running smoothly.

April 29, 2019, 09:11	starting cfx solver 19.2 on hpc	#1
Svetlana Senior Member Svetlana Tkachenko Join Date: Oct 2013 Location: Australia, Sydney Posts: 416 Rep Power: 15	When i start cfx5solve 19.2 on a cluster, it partitions the mesh and things, comes to '+--------------+ Solver +--------------+' and then stops there for many hours doing nothing. I am not at all sure how to troubleshoot it. Can it be a problem with the licence?

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
CFX Solver stopped with error when requested for backup during solver running	Mfaizan	CFX	40	May 13, 2016 07:50
thobois class engineTopoChangerMesh error	Peter_600	OpenFOAM	4	August 2, 2014 10:52
Working directory via command line	Luiz	CFX	4	March 6, 2011 21:02
why the solver reject it? Anyone with experience?	bearcat	CFX	6	April 28, 2008 15:08
CFX 5.5	Roued	CFX	1	October 2, 2001 17:49

April 29, 2019, 09:50		#2
Gert-Jan Senior Member Gert-Jan Join Date: Oct 2012 Location: Europe Posts: 1,928 Rep Power: 28	Stops for hours? Do you mean that it is doing something but that you don't know what it is doing, and then after several hours continues normally? E.g. have you included radiation? I have seen cases where it took hours for CFX to do the raytracing, but then continued normally....... Or is it really doing nothing and is waiting for you to kill the job? Does CFX leave a text file with an error in the directory that is named after the output file?

April 29, 2019, 10:43		#3
evcelica Senior Member Erik Join Date: Feb 2011 Location: Earth (Land portion) Posts: 1,188 Rep Power: 23	This could be the result of many things. I do not think it is the licence though, that would give you an error. Usually something is not right with the configuration of distributed CFX. You should run small test model distributed until you get everything set up correctly. Uninstall old versions or set the environmental variables correctly. Ensure "hosts" file is correct. Allow program and processes through firewall including MPI processes. perhaps start on only 2 machines distributed, as there may be a problem with only 1 of the nodes. This could be difficult to troubleshoot.

April 29, 2019, 17:56		#4
Svetlana Senior Member Svetlana Tkachenko Join Date: Oct 2013 Location: Australia, Sydney Posts: 416 Rep Power: 15	No radiation. It does not continue normally. Iteration number 1 does not begin. I'll test with smaller distributed runs and write back. Thanks!

May 8, 2019, 10:22		#5
kjetil Senior Member Join Date: Mar 2009 Location: Norway Posts: 138 Rep Power: 17	It is possibly a problem with IntelMPI. You might need to install the "old" version, which is found in the installation directory.

May 8, 2019, 22:02		#6
Svetlana Senior Member Svetlana Tkachenko Join Date: Oct 2013 Location: Australia, Sydney Posts: 416 Rep Power: 15	You're right, I was loading a conflicting version of openmpi in my bashrc because I needed this for another script. Now I can see how this was a confusing and poor place to put that line in. I removed it and everything started running smoothly.