code saturne 2.1.5-1 from debian sid repository

[eitriv]

Hello
Using Debian Sid installed Code Saturne from the repository.
The GUI is working well but when I start calculation I have this output in the terminal
Any idea how to solve this
Thanks

et@itx:~/saturnecase/tutor/tuto/DATA$ ./SaturneGUI
nohup: ignoring input and redirecting stderr to stdout
Traceback (most recent call last):
File "/usr/bin/code_saturne", line 49, in <module>
retcode = cs.execute()
File "/usr/lib/python2.7/dist-packages/code_saturne/cs_script.py", line 62, in execute
return self.commands[command](options)
File "/usr/lib/python2.7/dist-packages/code_saturne/cs_script.py", line 114, in run
return cs_run.main(options, self.package)
File "/usr/lib/python2.7/dist-packages/code_saturne/cs_run.py", line 201, in main
save_results=save_results)
File "/usr/lib/python2.7/dist-packages/code_saturne/cs_case.py", line 1702, in run
mpi_environment)
File "/usr/lib/python2.7/dist-packages/code_saturne/cs_case.py", line 1258, in prepare_data
n_procs_default)
File "/usr/lib/python2.7/dist-packages/code_saturne/cs_exec_environment.py", line 1198, in __init__
self.mpi_env = mpi_environment(pkg, self.resources, wdir)
File "/usr/lib/python2.7/dist-packages/code_saturne/cs_exec_environment.py", line 635, in __init__
init_method(p, resource_info, wdir)
File "/usr/lib/python2.7/dist-packages/code_saturne/cs_exec_environment.py", line 899, in __init_openmpi__
launcher_base = os.path.basename(self.mpiexec)
File "/usr/lib/python2.7/posixpath.py", line 112, in basename
i = p.rfind('/') + 1
AttributeError: 'NoneType' object has no attribute 'rfind'

[Yvan Fournier]

Hello,

I did not reproduce your error (on a virtual machine running SID, updated yesterday), but I recall having encountered a similar issue a few weeks ago.

If you have a mpiexec or mpiexec.mpi command in your search path, the issue should not occur, so I suspect you do not have the "mpi-default-bin" package installed.

If you do not want to install it, you may either add the path to OpenMPI's "mpiexec" in you PATH (for example in your .bashrc), or explicitely tell the code which mpiexec command it should use by using the matching portion in the cs_user_scripts.py in the DATA directory (copied from DATA/REFERENE).

Best regards,

Yvan

[eitriv]

Hello Yvan,
Thank you that partially solved the problem.
I have installed mpi-default-bin which took also openmpi.bin
now I can run the case with a single process but with two processes I get the error below.
You mentioned other solutions, is there a reason not to install mpi-default-bin?
Best regards

et@itx:~/saturnecase/tutor/tuto/DATA$ ./SaturneGUI
nohup: ignoring input and redirecting stderr to stdout
code_saturne is running
***********************
Version: 2.1.5
Path: /usr
Result directory:
/home/et/saturnecase/tutor/tuto/RESU/20120329-1300

Parallel code_saturne on 2 processes.

****************************
Preparing calculation data
****************************

***************************
Preprocessing calculation
***************************

**********************
Starting calculation
**********************

[itx:03480] [[INVALID],INVALID] ORTE_ERROR_LOG: Not found in file ../../../../../../orte/mca/ess/hnp/ess_hnp_module.c at line 161
--------------------------------------------------------------------------
It looks like orte_init failed for some reason; your parallel process is
likely to abort. There are many reasons that a parallel process can
fail during orte_init; some of which are due to configuration or
environment problems. This failure appears to be an internal failure;
here's some additional information (which may only be relevant to an
Open MPI developer):

orte_plm_base_select failed
--> Returned value Not found (-13) instead of ORTE_SUCCESS
--------------------------------------------------------------------------
[itx:03480] [[INVALID],INVALID] ORTE_ERROR_LOG: Not found in file ../../../orte/runtime/orte_init.c at line 132
--------------------------------------------------------------------------
It looks like orte_init failed for some reason; your parallel process is
likely to abort. There are many reasons that a parallel process can
fail during orte_init; some of which are due to configuration or
environment problems. This failure appears to be an internal failure;
here's some additional information (which may only be relevant to an
Open MPI developer):

orte_ess_set_name failed
--> Returned value Not found (-13) instead of ORTE_SUCCESS
--------------------------------------------------------------------------
[itx:03480] [[INVALID],INVALID] ORTE_ERROR_LOG: Not found in file ../../../../../orte/tools/orterun/orterun.c at line 543
solver script exited with status 243.

Error running the calculation.

Check code_saturne log (listing) and error* files for details.


****************************
Saving calculation results
****************************

Error in calculation stage.

[Yvan Fournier]

Hello,

I would suspect some conflict between multiple versions of OpenMPI, as Debian (and similar) distributions may have multiple versions for different dependencies.

When I checked (yesterday evening), if I rember correctly, the version of OpenMPI used was 1.4.3 (not 1.5).

I had issues with MPI a few weeks ago (I don't rember whether it was on my Debian SID virtual machine or when upgrading by main Ubuntu machine from 11.10 to 12.04-beta), but uninstalling all MPI packages and reinstalling those needed (to let the system re-sort dependencies) fixed the problem. Reinstalling the Code_Saturne package after that is probably a safe bet also.

Best regards,

Yvan

[eitriv]

Hello
I did not find any nmultiple of mpi, here is the output of: dpkg -l | grep mpi
Thanks for your help
Best regards,
Eitan

et@itx:~$ dpkg -l | grep mpi
ii g++ 4:4.6.3-3 GNU C++ compiler
ii g++-4.6 4.6.3-1 GNU C++ compiler
ii gcc 4:4.6.3-3 GNU C compiler
ii gcc-4.4-base:amd64 4.4.7-1 GCC, the GNU Compiler Collection (base package)
ii gcc-4.6 4.6.3-1 GNU C compiler
ii gcc-4.6-base:amd64 4.6.3-1 GCC, the GNU Compiler Collection (base package)
ii gcc-4.7-base:amd64 4.7.0-1 GCC, the GNU Compiler Collection (base package)
ii libhdf5-mpi-dev 1.8.8-9 Hierarchical Data Format 5 (HDF5) - development files - MPICH2 version
ii libhdf5-openmpi-7 1.8.8-9 Hierarchical Data Format 5 (HDF5) - runtime files - OpenMPI version
ii libhdf5-openmpi-dev 1.8.8-9 Hierarchical Data Format 5 (HDF5) - development files - OpenMPI version
ii libopenmpi-dev 1.4.3-2.1 high performance message passing library -- header files
ii libopenmpi1.3 1.4.3-2.1 high performance message passing library -- shared library
ii liborc-0.4-0 1:0.4.16-1 Library of Optimized Inner Loops Runtime Compiler
ii make 3.81-8.1 An utility for Directing compilation.
ii mpi-default-bin 1.0.1 Standard MPI runtime programs (metapackage)
ii mpi-default-dev 1.0.1 Standard MPI development files (metapackage)
ii openmpi-bin 1.4.3-2.1 high performance message passing library -- binaries
ii openmpi-common 1.4.3-2.1 high performance message passing library -- common files
et@itx:~$

[Yvan Fournier]

Hello,

What is the output of "type mpiexec" or "type mpiexec.openmpi" on your system ?
If none of those are found, try reinstalling the "mpi-default-bin" pacakge.

Best regards,

Yvan

[eitriv]

Hello,

both are found:

et@itx:~$ type mpiexec
mpiexec is /usr/bin/mpiexec
et@itx:~$ type mpiexec.openmpi
mpiexec.openmpi is /usr/bin/mpiexec.openmpi
et@itx:~$

Best regards,

Eitan

[Yvan Fournier]

Hello,

Sorry, coming back from the week-end, I was one step behind and still thinking about the previous issue, so my last suggested test was not very useful.

To test your OpenMPI installation, could you simply try to run:

mpiexec -n 2 /bin/hostname

to see if you have similar orte (Open MPI Runtime Environment) error messages ?

If this also fails, reinstalling/updating Open MPI packages might help (you are running an "unstable"/bleeding edge version of Debian after all).

Best regards,

Yvan

[eitriv]

Hello,

Yes, same orte (Open MPI Runtime Environment) error messages.

But re-installing all open-mpi packages did not help, maybe I will try to post on the open-mpi project forum.

Thanks anyway,

Best regards

Eitan

[Yvan Fournier]

Hello,

what is surprising is that I did not encounter this issue on a test (virtual) machine under Debian sid.
I suspect this is more of a Debian packaging (subtle conflicts between packages, or subdivision of a single install into several packages which are not as orthogonal as thought) than purely Open MPI issue, so you might also want to check the issue on the Debian side.

Best regards,

Yvan