CFD Online Discussion Forums

CFD Online Discussion Forums (
-   FLUENT (
-   -   Batch Mode (

Titiksh Patel September 6, 2004 11:11

Batch Mode
Hi Everyone

Did anybody have knowledge of working in batch mode systems (Connecting windows 2000 to Linux systems).

I am facing some problems in DPM model.

Regards, Titiksh

Tang September 7, 2004 11:49

Re: Batch Mode
Yes, I do. What kind of problems?

Titiksh Patel September 7, 2004 12:05

Re: Batch Mode
Hi Tang

Thanx for the reply. Actually I was trying to run FLUENT in parallel mode, but unsuccessfully.

I have mailed my details to your mail-id.

Regards, Titiksh

Titiksh Patel September 7, 2004 12:09

Re: Batch Mode
Hi Tang

I have been given an account in LINUX machine which I access through my PC (Windows2000) through putty.

I am attaching the .jou file and .pbs file. Kindly look into the files.

I am able to run serial jobs easily and getting results.

Actually I was not able to partition the mesh and was not able to locate different machines so that it can run parallel.

This is the journal file

rc 4litres_closed_0.77.cas rd 4litres_closed_02000.dat parallel/auto partition /file/autosave/data-frequency 1000 /solve/set/time-step 0.002 /solve/dual-time-iterate 8000 20 exit yes

---------------------------- This is pbs (parallel batch mode) file

#!/bin/sh #PBS-N fluent_run5 #PBS -l walltime=24:00:00 #PBS -q lp #PBS -l nodes=1:ppn=2 #PBS -l cput=48:00:00 echo "running on node " `cat $PBS_NODEFILE` /usr/support/Fluent.Inc/bin/fluent 3d -g t2 <4lc_1_0.77_pll.jou> 4lc_1_0.77_output_pll echo "done"

----------------------------- Regards, Titiksh

Tang September 7, 2004 19:00

Re: Batch Mode
Where I work we have a different method submitting jobs for batching. Are you able to monitor your job? We submit our jobs through a UNIX based station to a LINUX cluster. You can only partion the mesh in serial. I would suggest partioning your mesh in serial using the Metis method, this is the best method I have seen, and then saving your case and then resubmitting your case. This will eliminate the partion create in your journal file. Hope this helps.

alven299 August 26, 2009 01:35

Hi all,

I am a first time user of FLUENT running in cluster. I am operating with a UNIX based station to a LINUX cluster. Could you anyone please explain to me regarding:
-What is .jou and .pbs file?
-How do I create these files?
-Why .jou and .pbs instead of .cas?
-What are the command line for submitting cases to the cluster?

Since the UNIX based system involves no GUI, how do I define the type of solver, time-step, number of iteration, etc before running the case?

Thank you.

-mAx- August 26, 2009 07:44

.jou and . dbs are gambit file (Mesher)
Once your mesh is ready you have to export it as .msh file.
Then you import it in fluent and you can save your case file .cas , and also your data file .dat, once you finished to iterate.
Regarding you system without GUI, you can access to all pull down menu with TUI commands
Once fluent is loaded, press enter and you will see the menus
Check the Text Command List in the help

alven299 August 27, 2009 05:59

Thanks a lot Max for your explanation....:)

alven299 January 24, 2010 04:55

Hi there,

I am currently using an SSH client (Putty) to access a Unix based cluster. I have no problem in accessing the parallel FLUENT and so far have been successfully running my simulations, however, I have to stay connected to Putty i.e. keep my PC (remote computer) on during the iterations. If I disconnect the Putty, so does the iterations. This will certainly create a lot of inconveniences if my iterations require 3 months to converge.

Therefore, is there any method for me to submit the job to the cluster and off my PC (disconnect from the Putty) then come back the next three months to collect the data files? I suspect batch mode is the answer, but I am alien to this. Could someone help me by elaborating the term?

Thank you v much in advance.


-mAx- January 25, 2010 02:43

Is your remote PC considered as a Spawn Node?

alven299 January 25, 2010 02:53

I am not sure what does the term spawn node mean but this remote PC (windows OS) is used to access the Unix cluster via SSH client (Putty) and is not involved in the iterations work. Any clue? thank you:)

los January 28, 2010 07:57

Use the screen command, here is a tutorial:

alven299 February 1, 2010 10:40

Thank you very much los, your suggestion works wonderfully:) Unlike batch mode, this works easily without the need of script writing which is painful for me. Thank you!

los February 1, 2010 15:48

I am glad it worked =D

any further questions send me a message /o/

All times are GMT -4. The time now is 14:03.