Hi, has anyone done the parallel computations before? Can you give me some examples of the batch and journal files of running the fluent on the clusters?

I am using the clusters to run my fluent cases. But I got problems.

I used the batch file and the journal file. But always got errors.

Batch files ################################################# #PBS -N test_job #PBS -q cpu1 #PBS -l mem=3gb #PBS -l walltime=30:00 #PBS -m bea #PBS -M l.bai@swansea.ac.uk

fluent 2ddp -g -i /civeng/cvcluster09/cgpolynk/temp/temp > /civeng/cvcluster09/cgpolynk/tempexample.out #################################################

Journal file ################################################# rcd /civeng/cvcluster09/cgpolynk/temp/tube_pn=0n=1.cas it 500 wcd /civeng/cvcluster09/cgpolynk/temp/tube_pn=0n=1_%i.cas exit #################################################

Can you have a look on my procedure and give some suggestions how to solve this?

Any suggestion are appreciated.

Cheers.

could you please specify what is the problem? what error do you get? I'm using the following script and journal file:

#!/bin/bash

#PBS -l walltime=18:00:00

#PBS -l nodes=2pn=4

#PBS -M bogdan.nichita@epfl.ch

#PBS -m bae

#PBS -V

#

cd $PBS_O_WORKDIR

time fluent 3ddp -pnet -t8 -cnf=$PBS_NODEFILE -gu

< fluent_unsteady > microchannel_05_20.txt

#

rcd microchannel_05_20_2bubbles.cas.gz

solve set time-step 0.000001

solve dual-time-iterate

12000

20

exit

y

Thank you for sharing files.

My problem is that:

I can start the fluent using script, but it can not find the journal file in the batch command. I qsubed the job under my working directory, and the script and journal are in the same fold. But it can not recognize the journal.

I use qsub in my working directory. So I dont think I should include the full path in the script and journal file.

Is that batch command right?

cd /civeng/cvcluster09/cgpolynk/temp fluent 2ddp -g < temp >& results.txt &

or should I ignore the cd command?

there are some variables available within PBS, one of them is $PBS_O_WORKDIR which contains the path to your working directory, so before writing the fluent command just put

" cd $PBS_O_WORKDIR"

when you submit a job from your working directory, and the job starts, you are in your home directory and not in your working directory.

I am trying to using this script:

#!/bin/bash

#PBS -N test_job

#PBS -q cpu1

#PBS -l mem=3gb

#PBS -l walltime=30:00

#PBS -m bea

#PBS -M l.bai@swansea.ac.uk

#PBS -V

cd $PBS_O_WORKDIR

fluent 2ddp -g < jou > result.txt

the errors are

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

stty: standard input: Invalid argument
:
: No such file or directory

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Hi, I combined the script and journal file into one script file, and ran it. It launched fluent, but it didnt recognize the commands like rcd, it, wcd.

#PBS -N test_job

#PBS -q cpu1

#PBS -l mem=3gb

#PBS -l walltime=30:00

#PBS -m bea

#PBS -M l.bai@swan.ac.uk

fluent 2ddp -g > temp1.out

rcd

tube_pn=0n=1.cas

it

500

wcd

tube_pn=0n=1_%i.cas

exit

yes

The output file is: ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

stty: standard input: Invalid argument stty: standard input: Invalid argument

: Command not found. /civeng/cvcluster09/cgpolynk/Fluent.Inc/fluent6.2.16/cortex/lnamd64/cortex.3.6.6 -f fluent -g (fluent "2ddp -path/civeng/cvcluster09/cgpolynk/Fluent.Inc") Starting exec /civeng/cvcluster09/cgpolynk/Fluent.Inc/fluent6.2.16/lnamd64/2ddp/fluent.6.2.16 -cx node8:33218:33219

Welcome to Fluent 6.2.16

Copyright 2005 Fluent Inc.

All Rights Reserved

Loading "/civeng/cvcluster09/cgpolynk/Fluent.Inc/fluent6.2.16/lib/flprim.dmp.1119-64" Done.

rcd
: Command not found. tube_pn=0n=1.cas
: Command not found. it
: Command not found. 500
: Command not found. wcd
: Command not found. tube_pn=0n=1_%i.cas
: Command not found. exit
: Command not found. yes
: Command not found.

are you running on Linux?

did you try running one calculation on your cluster without batch? For starting the parallel process you should type a specific arguments which depends on your architecture. for example> fluent 3ddp -pnet

try like this:

#PBS -N test_job

#PBS -q cpu1

#PBS -l mem=3gb

#PBS -l walltime=30:00

#PBS -m bea

#PBS -M l.bai@swan.ac.uk

fluent 2ddp -g <<__EOF__> temp1.out

rcd

tube_pn=0n=1.cas

it

500

wcd

tube_pn=0n=1_%i.cas

exit

yes

__EOF__

yes, linux on clusters

yes, I tried, it works. I started it using

fluent 2ddp -g rec ?????.cas it 500 wcd ?????.cas yes exit

These command succeeded in the SSH by typing these commands.

are you sure you are running with parallel process? the -g argument means only that fluent won't execute the gui interface (ie: just in background) How many CPU do you have in your cluster? (check the number of partitions in your cas file. this number should be the same as the spawned nodes)

yes, I am using SSH secure shell to log onto the Linux system, and running the parallel fluent on clusters.

It seems the parallel fluent doesnot know the command 'rcd', so I change to use this '/file/rcd'. It seems running now.

Thank you for your suggestions.

No, it still crashed. Not know the command /file/rcd.

I am also running parallel computation under linux cluster (RSH). can you copy and paste all the lines wich appears in your fluent window when you start fluent with your command, till it crashes.

#PBS -N test_job

#PBS -q cpu1

#PBS -l mem=3gb

#PBS -l walltime=30:00

#PBS -m bea

#PBS -M l.bai@swansea.ac.uk

fluent 2ddp -g > temp1.out

/file/rcd tube_pn=0n=1.cas

/file/autosave/data-frequency

1000

/solve/set/variable-time-stepping

yes # variable time stepping #

2 # global courant number #

0.009

# Truncation Error Tolerance [0.0099999998] #

1000 # Ending Time (s) [1000] #

0.0001 # Minimum Time Step (s) [9.9999997e-05] #

10 # Maximum Time Step (s) [10] #

0.5 # Minimum Step Change Factor [0.5]#

5 # Maximum Step Change Factor [5] #

1 # Number of Fixed Time Steps [1] #

/file/wcd tube_pn=0n=1_%i.cas

/exit

y

stty: standard input: Invalid argument stty: standard input: Invalid argument

: Command not found. /civeng/cvcluster09/cgpolynk/Fluent.Inc/fluent6.2.16/cortex/lnamd64/cortex.3.6.6 -f fluent -g (fluent "2ddp -path/civeng/cvcluster09/cgpolynk/Fluent.Inc") Starting exec /civeng/cvcluster09/cgpolynk/Fluent.Inc/fluent6.2.16/lnamd64/2ddp/fluent.6.2.16 -cx node8:33335:33336

Welcome to Fluent 6.2.16

Copyright 2005 Fluent Inc.

All Rights Reserved

Loading "/civeng/cvcluster09/cgpolynk/Fluent.Inc/fluent6.2.16/lib/flprim.dmp.1119-64" Done. /file/rcd: Command not found. /file/autosave/data-frequency
: Command not found. 1000
: Command not found. /solve/set/variable-time-stepping: Command not found. yes: standard output: No space left on device yes: write error 2: Command not found. 0.009: Command not found. 1000: Command not found. 0.0001: Command not found. 10: Command not found. 0.5: Command not found. 5: Command not found. 1: Command not found. /file/wcd: Command not found. /exit
: Command not found. y
: Command not found.

: Command not found.

: Command not found.

: Command not found.

: Command not found.

sorry for insisting, but are you sure you are running parallel version of fluent?

*start fluent on your cluster WITHOUT the -g argument and WITHOUT your journal file. you will see fluent GUI loading. Then go and check your nodes spawned in your cluster (Parallel>Network>Configure ) if you your nodes in the spawned nodes colums, then ok fluent parallel is running

*file/read/journal and load your journal file. this is an alternative for running batch.

do you still have error?