Running parallel calculation from Terminal

[So_nuc]

Hello,

I have tried to run a case (with internal coupling) in parallel mode using the version 5.0.10 of Code_Saturne by means of the command code_saturne --nprocs n. However it crashes and I do not really see the reason why. Does the parallel mode depend on the number of meshes for example ?

Thank you in advance for your help.

Best regards,

So_nuc

[Mohammad]

Hello,

If you want to run the code with x parallel cores, you have to use the following command:

Code: Select all

code_saturne run -n x

Example
For running the code with 4 parallel cores:

Code: Select all

code_saturne run -n 4

Regards,

Mohammad

[So_nuc]

Hello Mohammad,

Thank you for your reply. However I get the same error with your command:

Code: Select all

--------------------------------------------------------------------------
MPI_ABORT was invoked on rank 0 in communicator MPI_COMM_WORLD 
with errorcode 1.

NOTE: invoking MPI_ABORT causes Open MPI to kill all MPI processes.
You may or may not see output from other processes, depending on
exactly when Open MPI kills them.
--------------------------------------------------------------------------
--------------------------------------------------------------------------
mpiexec.openmpi noticed that process rank 2 with PID 4828 on node is234230 exited on signal 6 (Aborted).
--------------------------------------------------------------------------

Thanks in advance for the support if you have any solution.

Regards,

So_nuc

[Yvan Fournier]

Hello,

Does you case run in serial mode ?

The important information is probably in the part of the logs you did not paste.

Regards,

Yvan

[So_nuc]

Hello Yvan,

Yes it runs in serial mode. Unfortunately the calculation crashes in case of n > 2 processors without error log. Moreover the listing file does not seem to contain some information about the error.

Thank you for your attention.

Kind regards,

So_nuc

[Yvan Fournier]

Hello,

I assume you also have user functions.
They might be the cause of the crash.

Regards,

Yvan

[So_nuc]

Hello Yvan,

Thank you for this feedback. Actually, using the same source files, I get no problem with a geometry in parallel mode whatever the number of processors is. But with another geometry, it works in parallel mode until a certain number and above this numer, I get an error of boundary conditions:

Code: Select all

@ ALE METHOD
@
@ Inconsistency in boundary condition types
@   for mesh velocity.
@   (cf. message(s) above)
@
@ Check boundary conditions in the GUI or in usalcl.f90.
@

Première face avec erreur de définition de conditions aux limites
  (sur 2)
  a boundary condition type 0, centre (0, 0, 0)

That is weird because there is in that geometry more than 2 boundary faces. Maybe is it a generic message for a certain type of error ? I would suppose that the splitting of the geometry into several regions for parallel calculation in that case may not work.

Thank you for your attention.

Best regards,

So_nuc

[So_nuc]

Hello again,

I join you here the MPI rank IDs within the same geometry for 3 and 4 processors. In the last case, it seems that the geometry is not symmetrically cut which may explain the problem.

Thank you in advance for your attention.

Best regards,

So_nuc

[Yvan Fournier]

Hello,

Do you have user source files or not for this case ? If you do, the error might be in those files, so can you post them ? Otherwise, there may be a bug, so the test case would be good.

Regards,

Yvan