[Error] Fluent Not Responding on Workbench Environment

Luiz · August 27, 2018, 16:09

Hello,

I've been working in Fluent for the past months and there is one recurrent error that happens when simulating through workbench on Linux (Ansys 17.0 on CentOS). Depending on the number of iterations, from 15 to 25K depending on the mesh, the program ceases to respond without returning any error or message. The only thing that pops up is the system message.
Some few hypothesis came in mind, one that fluent requires an special permission for writing files and the second one that it just takes some time to unload the memory, since it still uses a certain percentage of RAM available. I can't open Fluent and simulate from it, due to the lack of GPU from the server we use at the facility.

Did anyone come across this error or has a suggestion of what to do?

Many thanks,
Luiz

LuckyTran · August 27, 2018, 16:50

There are no special permissions in unix/linux, only permissions.

Why don't you try running Fluent without the gui to debug? You can launch Fluent in nogui mode...

Luiz · August 27, 2018, 17:07

I didn't open in the no gui mode (trying this rn), because I was using Solution Monitoring to oversee the simulation. Even after the program crashes, I can see the transcript from fluent and it doesn't show any error, only that one of the processes were killed.
The reason I killed the process is that the CPU usage went to zero and there was no updates on the monitors anymore.

blackmask · August 27, 2018, 22:15

Issue

Code:

dmesg

in the terminal to check whether this is an out-of-memory issue. Did you run fluent in parallel across multiple nodes or in a single node? Did you use any UDF in your calculation?

Luiz · August 28, 2018, 00:01

I am not using any UDF, and the parallel calculus uses 52/64 cores in a single machine. I will take a look at this issue, but it doesn't seem to be the case since RAM usage doesn't go above 19/128 Gb. I am not very familiar with this term, but this error can be that the process needs more memory than what CentOS allows right? If it is the case how can I improve this limit? Increasing its swap?

blackmask · August 28, 2018, 02:47

Will if you have 128 Gb of RAM then you might not need swap space at all. You can check various settings by

Code:

ulimit -a

to see anything abnormal.

You really should launch fluent in console mode (i.e., without GUI) so that you could capture the error message. Once you have solved the problem you can run in graphical mode as you like.

Luiz · August 31, 2018, 18:24

I managed to simulate it further than before without any error without GUI. Although, I encountered a difficulty when it comes to export the data within a time range, ex: data in 1 s, 2 s, 3s and the final solution.
I wrote a journal script, although I am not certain what expression to use to export these results to CFD-Post. Its idea is quite simple, just do an permanent solution and use it as an initialization to a transient one.

Code:

rc Trans-sst-init.cas
solve/initialize/hyb-initialization
it 50000
wd

rc LES-200.cas
yes
rd Trans-sst-init.dat
file/auto-save/data-frequency 14104
solve/dual-time-iterate 101373
wd LES-200.dat
/file/export/cfd-post-compatible

Many thanks!

blackmask · August 31, 2018, 23:33

You can set it from GUI under [file]->export->during calculation.

Luiz · September 5, 2018, 22:55

Does it keep this configuration once I run no GUI?
Also, for when it runs in transient mode in GUI I can save my time series in one file using autosave(timesteps), the export process goes through workbench. Is it possible to use this same autosave and export a time series with [file]->export->cfd post compatible in a same file using a journal?
If it doesn't should I write an UDF for such?

Thanks in advance.

August 27, 2018, 16:09	[Error] Fluent Not Responding on Workbench Environment	#1
Luiz New Member Luiz Oliveira Join Date: Aug 2018 Location: Napoli, Italy Posts: 24 Rep Power: 8	Hello, I've been working in Fluent for the past months and there is one recurrent error that happens when simulating through workbench on Linux (Ansys 17.0 on CentOS). Depending on the number of iterations, from 15 to 25K depending on the mesh, the program ceases to respond without returning any error or message. The only thing that pops up is the system message. Some few hypothesis came in mind, one that fluent requires an special permission for writing files and the second one that it just takes some time to unload the memory, since it still uses a certain percentage of RAM available. I can't open Fluent and simulate from it, due to the lack of GPU from the server we use at the facility. Did anyone come across this error or has a suggestion of what to do? Many thanks, Luiz

August 27, 2018, 22:15		#4
blackmask Senior Member Join Date: Aug 2011 Posts: 421 Blog Entries: 1 Rep Power: 22	Issue Code: dmesg in the terminal to check whether this is an out-of-memory issue. Did you run fluent in parallel across multiple nodes or in a single node? Did you use any UDF in your calculation?

August 28, 2018, 02:47		#6
blackmask Senior Member Join Date: Aug 2011 Posts: 421 Blog Entries: 1 Rep Power: 22	Will if you have 128 Gb of RAM then you might not need swap space at all. You can check various settings by Code: ulimit -a to see anything abnormal. You really should launch fluent in console mode (i.e., without GUI) so that you could capture the error message. Once you have solved the problem you can run in graphical mode as you like.

August 31, 2018, 18:24		#7
Luiz New Member Luiz Oliveira Join Date: Aug 2018 Location: Napoli, Italy Posts: 24 Rep Power: 8	I managed to simulate it further than before without any error without GUI. Although, I encountered a difficulty when it comes to export the data within a time range, ex: data in 1 s, 2 s, 3s and the final solution. I wrote a journal script, although I am not certain what expression to use to export these results to CFD-Post. Its idea is quite simple, just do an permanent solution and use it as an initialization to a transient one. Code: rc Trans-sst-init.cas solve/initialize/hyb-initialization it 50000 wd rc LES-200.cas yes rd Trans-sst-init.dat file/auto-save/data-frequency 14104 solve/dual-time-iterate 101373 wd LES-200.dat /file/export/cfd-post-compatible Many thanks!

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Workbench 18.2 Installation Error Ubuntu 16.04	turbostudent	ANSYS	1	October 3, 2019 11:13
Fluent parametric analysis with workbench runs solwly	hjxy2012	FLUENT	1	August 28, 2016 00:16
Workbench Error in fluent could not handle event save project	pradyuman	FLUENT	0	May 26, 2016 04:17
Get Maxwell ansoft results in ansys workbench and find the flow velocity in Fluent, i	rrahman	FLUENT	0	April 16, 2016 22:11
The fluent stopped and errors with "Emergency: received SIGHUP signal"	yuyuxuan	FLUENT	0	December 3, 2013 23:56

August 27, 2018, 16:50		#2
LuckyTran Senior Member Lucky Join Date: Apr 2011 Location: Orlando, FL USA Posts: 5,761 Rep Power: 66	There are no special permissions in unix/linux, only permissions. Why don't you try running Fluent without the gui to debug? You can launch Fluent in nogui mode...

August 27, 2018, 17:07		#3
Luiz New Member Luiz Oliveira Join Date: Aug 2018 Location: Napoli, Italy Posts: 24 Rep Power: 8	I didn't open in the no gui mode (trying this rn), because I was using Solution Monitoring to oversee the simulation. Even after the program crashes, I can see the transcript from fluent and it doesn't show any error, only that one of the processes were killed. The reason I killed the process is that the CPU usage went to zero and there was no updates on the monitors anymore.

August 28, 2018, 00:01		#5
Luiz New Member Luiz Oliveira Join Date: Aug 2018 Location: Napoli, Italy Posts: 24 Rep Power: 8	I am not using any UDF, and the parallel calculus uses 52/64 cores in a single machine. I will take a look at this issue, but it doesn't seem to be the case since RAM usage doesn't go above 19/128 Gb. I am not very familiar with this term, but this error can be that the process needs more memory than what CentOS allows right? If it is the case how can I improve this limit? Increasing its swap?

August 31, 2018, 23:33		#8
blackmask Senior Member Join Date: Aug 2011 Posts: 421 Blog Entries: 1 Rep Power: 22	You can set it from GUI under [file]->export->during calculation.

September 5, 2018, 22:55		#9
Luiz New Member Luiz Oliveira Join Date: Aug 2018 Location: Napoli, Italy Posts: 24 Rep Power: 8	Does it keep this configuration once I run no GUI? Also, for when it runs in transient mode in GUI I can save my time series in one file using autosave(timesteps), the export process goes through workbench. Is it possible to use this same autosave and export a time series with [file]->export->cfd post compatible in a same file using a journal? If it doesn't should I write an UDF for such? Thanks in advance.