Coupling in code_saturne

[Mahmouddwikat]

Hi,

I was trying to couple between different meshes (rotor and stator) in Code_saturne v3.0.1 and I use runcase_coupling to run the simulation, and it gave me the following error.
TypeError: coercing to Unicode: need string or buffer, NoneType found

"
Traceback (most recent call last):
File "./runcase_coupling", line 172, in <module>
save_results=True)
File "/opt/Code_saturne/v3.0.1/code_saturne-3.0.1/lib/python2.7/site-packages/code_saturne/cs_case.py", line 1535, in run
mpi_environment)
File "/opt/Code_saturne/v3.0.1/code_saturne-3.0.1/lib/python2.7/site-packages/code_saturne/cs_case.py", line 1144, in prepare_data
self.set_result_dir(self.run_id)
File "/opt/Code_saturne/v3.0.1/code_saturne-3.0.1/lib/python2.7/site-packages/code_saturne/cs_case.py", line 453, in set_result_dir
if not os.path.isdir(self.result_dir):
File "/usr/lib64/python2.7/genericpath.py", line 41, in isdir
st = os.stat(s)"

Any help in this issue

Thanks

[Yvan Fournier]

Hello,

Are you using the GUI ? Did you define a scratchdir in code_saturne.cfg ?

Regards,

Yvan

[Mahmouddwikat]

Hey Yvan,

I don't use GUI.

Yes, I defined it in runcas_coupling.
and I defined it as well in each "cs_user_scripts.py" for each case.

Thanks

Hello,

Are you using the GUI ? Did you define a scratchdir in code_saturne.cfg ?

Regards,

Yvan

[Yvan Fournier]

Hello,

Could you post your case's setup (everything in each case's DATA diectory, as well as the runcase_coupling; I won't need the rest to test such an early crash).

Regards,

Yvan

[Mahmouddwikat]

Hey,

Yes sure.



Thanks

Hello,

Could you post your case's setup (everything in each case's DATA diectory, as well as the runcase_coupling; I won't need the rest to test such an early crash).

Regards,

Yvan

[Yvan Fournier]

Hello,

Do you have a RESU_COUPLING directory at the same level as runcase_coupling ?

I can reproduce your issue only when this directory is missing (I'll ensure the error message is more explicit in this case in future releases).

Also, if you do not want a separate scratch directory, defining SCRATCHDIR is not useful. Just set it to None (or to '/', as all directories are a subdirectory of the latter, and the code avoids using a separate scratch directory if a case is already in the scratch directory).

Regards,

Yvan

[Mahmouddwikat]

Hey Yvan,

Many thanks for your help.
My mistake was about not creating the case together which is not produce the RESU_COUPLING.

I pass that error and now I have something related to "run_solver.sh"
But I'm trying to check the following modules which can't be found
line 10: module: command not found
line 11: module: command not found

"
module purge
module load mpi/gcc/openmpi/1.6-ib
"

[Yvan Fournier]

Hello,

You can ignore the "module not found" warning. The run_solver script tries to reload/unload modules in case it is launched alone (after a "code_saturne run --initialize" for example), but when launched from the runcase_coupling script, the "module" alias is not found (which is OK as the runcase_coupling also sets the corrrect modules before calling run_solver).

We'll have to improve this, but this is usually not a big issue even if the message is a bit alarming.

Best regards,

Yvan

[Mahmouddwikat]

Hello,

Is it possible to run my coupled test case on one processor?
Its small mesh, just to test my modified subroutine.

Thanks

[Yvan Fournier]

Hello,

You will need at least 1 "logical" processor (MPI rank) per domain, but there is no problem with running it on a single "physical" processor on a workstation. Peformance won't be optimal, but if it's just for testing, you probably don't care.

The only cases where MPI might complain about "oversubscription" is when integration with a batch system/resource manager is setup to avoid this, for performance reasons. On a workstation with no batch system, things will be fine.

Best regards,

Yvan Fournier