Problem in

[Renan Neves Nascimento]

Hi everybody,
 
I'm having a problem when I try to check my mesh. The error that appears is this:
 
p, li { white-space: pre-wrap; }
Lecture du fichier de maillage au format MED (EDF/CEA)
------------------------------
Fichier de maillage : /home/renanneves/CYL/MESH/Mesh_1.med
Nom du maillage : Mesh_1
Erreur a l'exécution de l'ECS
=============================
ecs_pre_med.c:427: Erreur fatale.
MED : erreur à la lecture du fichier "/home/renanneves/CYL/MESH/Mesh_1.med".
Erreur à la lecture des coordonnées.
Pile d'appels
1: 0x7f19258b4152 (libbft.so.1)
2: 0x427b13 (cs_preprocess)
3: 0x40d51b (cs_preprocess)
4: 0x409570 (cs_preprocess)
5: 0x7f19248cbc4d (libc.so.6)
6: 0x403839 (cs_preprocess)
Fin de la pile



____________________
 
In attach there's my mesh.
 
By the way, I tried to save as UNV. The check part worked, but when I tried to run appeared another error:
 
  ********************************************
             Starting calculation
  ********************************************
 
mpiexec_ubuntu: cannot connect to local mpd (/tmp/mpd2.console_renanneves); possible causes:
  1. no mpd is running on this host
  2. an mpd is running but was started without a "console" (-n option)
In case 1, you can start an mpd on this host with:
    mpd &
and you will be able to run jobs just on this host.
For more details on starting mpds on a set of hosts, see
the MPICH2 Installation Guide.
Error running the calculation.
 
Check Kernel log (listing) and error* files for details
 
 
  ********************************************
         Error in calculation stage.
  ********************************************
 
 
_____
_________
 
 
And I would like to know too, what is the difference between UNV's and MED's files.
 
 
 
 
Really thanks for all in advance!

[Yvan Fournier]

Hello,

For the probem with MED files, you probably build the Code_Saturne preprocessor (ecs module) with MED 2.3.6, but SALOME 6.3 outputs files in MED 3.0 format. Version 2.0.2 of the preprocessor may support either MED 2.3.6 and 3.0, depending on the version you indicate when building the code (older versions only support version 2.3.6).

So the best solution is to reinstall the Code_Saturne preprocessor with support for MED 2.3.6 (--with-med=<...> option in Preprocessor's configure option), or use the installer to install MED (the installer installs MED 3.0.3 if required, but may use any MED 2.3 or 3.0 version installed on your machine if you configure the setup to use MED, but not install it).

UNV file format is a common, but quite old file format. It does not require any additional library, which is an advantage, but it also does not support polygon, pyramid, or polyhedron element types, which SALOME can produce, and MED can handle, so it is really a last resort, and not my favorite solution at all.

In the UNV case, it seems your mesh was read correctly, but the code could not start because you tried to use multiple processors, but MPI (MPICH or MPICH2 in you case) is not configured correctly on your machine, and does not start correctly. You should not have any problem when running with only 1 processor. To run with multiple processors this, either configure MPICH2 by following its instructions (or finding colleagues or administrators who could help you with this, as this is not a Code_Saturne issue, but could affect other tools using MPI), or try OpenMPI to see if its default configuration works better for you (depending on your Linux distribution, either OpenMPI or MPICH2 way work better "out of the box", or both may work fine; both work fine with no additional configuration on my Ubuntu 11.10 install). If you change MPI version, though you need to reinstall/recompile Code_Saturne with the new distribution.

Regards,

Yvan

[Renan Neves Nascimento]

Hi,

I tried to use installer of edf's site, but it's not working:



Installation of Code_Saturne
____________________________

The process will take several minutes.
You can have a look at the log file meanwhile.

Check the setup file and some utilities presence.

o Checking for python... /usr/bin/python
Python version is 2.6
o Checking for make... /usr/bin/make
o Checking for pdflatex... /usr/bin/pdflatex
o Checking for fig2dev... /usr/bin/fig2dev

Installation of HDF5
hdf5-1.8.7.tar.gz is not a tar archive


And it finishes here.

Regards,

Renan

[Yvan Fournier]

Hello,
In the setup.py file, try to find the string: archive="hdf5-1.8.7.tar.gz" and replace 1.8.7 by 1.8.8 to see if this improves things (there is a new version of HDF5 available, so the old path may not be valid anymore).
Otherwise, on a distribution such as Debian or Ubuntu, you can install MED 3 with the Code_Saturne installer, using an existing (package) HDF5 install (you need to install hdf5-dev also).
Regards,
  Yvan

[Renan Neves Nascimento]

Hi.

My ".med" problem is solved. Now, the problem is with the post-processing.

That's the problem:




In the terminal

--------------------------------------------------------------------------
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.
--------------------------------------------------------------------------
--------------------------------------------------------------------------
mpirun has exited due to process rank 0 with PID 9718 on
node ubuntu exiting without calling "finalize". This may
have caused other processes in the application to be
terminated by signals sent by mpirun (as reported here).
--------------------------------------------------------------------------
Error running the calculation.

Check Kernel log (listing) and error* files for detail



and

In the file





fvm_writer.c:663: Erreur fatale.

Format type "MED_fichier " required for case "chr" is not available


Pile d'appels :
1: 0x7f1680323920 (libfvm.so.0)
2: 0x7f168067c67a (libsaturne.so.0)
3: 0x7f168067caf0 (libsaturne.so.0)
4: 0x7f168055a4f4 (libsaturne.so.0)
5: 0x7f168055af35 (libsaturne.so.0)
6: 0x7f167d63ac4d <__libc_start_main+0xfd> (libc.so.6)
7: 0x400759 <> (cs_solver)
Fin de la pile


Really thanks

Best Regards!

[Yvan Fournier]

Hello,

It seems that you have MED support for the preprocessor, but not for the postprocessing output (based on the configuration of the FVM library, whereas preprocessing is done by ECS).

Do you have multiple installs, or dit you run a manual or partially manual install ? If this is not the case, it might be possible that your installed HDF5 library is built with MPI, but a conflicting library compared to FVM / Code_Saturne (for example MPICH2 for one, OpenMPI for the other, either way).

If your HDF5 library uses MPI, using an HDF5 package not using MPI would avoid issues (Code_Saturne does not yet support Parallel CGNS or Parallel MED output anyways). After switching to such a package, the FVM library from Code_Saturne needs to be re-configured/reinstalled (as by default shared libraries are enabled, this is enough).

If you have multiple installs, removing them and doing a clean install should solve the issue.

Best regards,

  Yvan

[Renan Neves Nascimento]

I tried, but it changed nothing.
So, I was trying to run with the follow configuration:
 
Reference Step Time: 0.05
Number of Iterations: 300
Post-Processing every "n" steps: 3
 
And it didn't work.
 
But when I tried to run like this:
Reference Step Time: 0.1
Number of Iterations: 10
Post-Processing every "n" steps: 1
 
It worked! But it gave me nothing.
 
These were the only changes that I did, and gave a result. Obviously the problem is really in the "post-processing". Would you know what else I can do?
 
Best Regards
 

[Renan Neves Nascimento]

Could be another mesh problem?