Hi Fluent Users, sometimes I run Fluent in a batch job, i.e. I use the ability of Fluent to read journal or log files instead of interactive commands by the 'i-' option, the option of switching off gaphics '-g' and the display diversion function ('>') of UNIX. This gives me the opportunity to have a look on the progress of the calculation from a remote login.

Unfortunately it happens sometimes that the number of iterations was set to high in order to have finished next morning. This leads to my question:

Does someone know how to stop a fluent process / fluent batch job AND to save the data, i.e. not to loose the results of one night or a weekend? Sometimes kill -1 on the cortex process works. Fluent then writes the case as #FLUENT# and the data as #FLUENT#.dat.

Thank you already for answering.

Hi,

In the journal file you can add statements such that the fluent files are saved after a specified number of iterations after which iterations are resumed again. Example of a .jou file to run test.cas two times with 20 iterations each time.

file

read-case-data

test.cas

/solve

iterate

20

quit

/file

write-case-data

y

/solve

iterate

20

quit

/file

write-case-data

y

quit

quit

exit

This saves the case after 20 iterations and resumes iterating again. So if you need to run more iterations change the number of iterations or save and iterate a couple of times by copying the portion and pasting where necessary.

iterate

20

quit

/file

write-case-data

y where required.

Chetan

To check your job (i.e. to save and continue the run) create a file named check-fluent in your /tmp directory.

To save and exit, create a file named exit-fluent .

This should work if you don't run through any queue system. Queue systems normally have the check-batch-job option and this is what I have to use right now.

Anders

How do I have to create those files?

Use the command touch exit-fluent which creates a file named exit-fluent without any content.

Where does fluent save the data+case file? Into my tmp-directory?

No answers necessary, I found it out by myself: Inside the working directory form where the job was submitted a dat and cas file whose name includes the actual number of iterations is saved.

Very nice! Thanks a lot Anders Jerhamre.

[Svetlana]

need to save data and stop fluent; autosave is not enabled. looking for guides / tutorials

[Carles_SM]

Dear experts,

Submitting FLUENT 19.1 jobs on PBS qeueing system with scripts.
Is there a way to stop ongoing simulation and save .dat file?

I have tried creating a file in the same folder as the out.log file gets updated.
File names used:
exit-fluent (as suggested in http://jullio.pe.kr/fluent6.1/help/html/ug/node62.htm)
checkpoint-exit (but I think this works for SGE queueing system)

but job still runs...

Any ideas?

Regards,

[LuckyTran]

The check-fluent trick should work still on PBS, it's not SGE specific. Make sure you put it in the right place though. Are you able to find the cleanup file in the same dir?


If you are running Fluent thru workbench, the location might be different (where the .set file is).

[Carles_SM]

LuckyTran,

Working outside of Workbench.
Job is splitted on several compute nodes.
I create the files (tried with check-fluent and exit-fluent) where a file named
cleanup-fluent-***.sh is written and where I am able to follow the out.log evolution...

[Antech]

I usually enable periodic auto-save with my custom Fluent start script (we use Torque queue system so it's too inconvenient to launch the job manually every time). But, when the calculation is slow (i.e. skeletal/detailed chemistry), it would be better to save-and-exit after any iteration. I tried to create an empty exit-fluent file in run directory (where the case and data files are) without any effect (I checked that exit-fluent file is there and has correct name with file manager). Fluent version is 16.2. So it looks like this trick will only work for particular Fluent versions.

[Svetlana]

Originally Posted by Antech

I usually enable periodic auto-save with my custom Fluent start script (we use Torque queue system so it's too inconvenient to launch the job manually every time). But, when the calculation is slow (i.e. skeletal/detailed chemistry), it would be better to save-and-exit after any iteration. I tried to create an empty exit-fluent file in run directory (where the case and data files are) without any effect (I checked that exit-fluent file is there and has correct name with file manager). Fluent version is 16.2. So it looks like this trick will only work for particular Fluent versions.

Thought it's a file in /tmp, not in run directory

[Antech]

Thanks, touch /tmp/exit-fluent did the trick!