[SOLVED]CdA Parallel Version 12.5 does not find HDF5 and MED libraries

Viktor.Vlaski · 2016-03-22 07:45:56

Hello,

CdA, both sequential and parallel versions are compiled on a RedHat cluster without errors. During the configuration for both versions it was displayed that HDF5 and MED libraries have been found.

Execution of sequential version, runs without any problems:
-- CODE_ASTER -- VERSION : EXPLOITATION (stable) --

Version 12.5.0 modifi?e le 12/12/2015
r?vision ae36b60ebe4a - branche 'v12'
Copyright EDF R&D 1991 - 2016

Ex?cution du : Tue Mar 22 07:10:27 2016
Nom de la machine : frontep2
Architecture : 64bit
Type de processeur : x86_64
Syst?me d'exploitation : Linux Red Hat Enterprise Linux Server 6.7 Santiago 2.6.32-504.23.4.el6.x86_64
Langue des messages : (ANSI_X3.4-1968)

Version de Python : 2.7.3
Version de NumPy : 1.8.2
Parall?lisme MPI : inactif
Parall?lisme OpenMP : actif
Nombre de processus utilis?s : 4
Version de la librairie HDF5 : 1.8.10
Version de la librairie MED : 3.0.8
Librairie MFront : non disponible
Version de la librairie MUMPS : 4.10.0
Librairie PETSc : non disponible
Version de la librairie SCOTCH : 5.1.10

M?moire limite pour l'ex?cution : 3500.00 Mo
consomm?e par l'initialisation : 420.26 Mo
par les objets du jeu de commandes : 0.00 Mo
reste pour l'allocation dynamique : 3078.20 Mo
Taille limite des fichiers d'?change : 12.00 Go

When the parallel version is executed, HDF5 and MED are marked as not present !?:

-- CODE_ASTER -- VERSION : EXPLOITATION (stable) --

Version 12.5.0 modifi?e le 12/12/2015
r?vision ae36b60ebe4a - branche 'v12'
Copyright EDF R&D 1991 - 2016

Ex?cution du : Tue Mar 22 07:24:44 2016
Nom de la machine : frontep2
Architecture : 64bit
Type de processeur : x86_64
Syst?me d'exploitation : Linux Red Hat Enterprise Linux Server 6.7 Santiago 2.6.32-504.23.4.el6.x86_64
Langue des messages : (ANSI_X3.4-1968)

Version de Python : 2.7.3
Version de NumPy : 1.8.2
Parall?lisme MPI : actif
Rang du processeur courant : 0
Nombre de processeurs utilis?s : 1
Parall?lisme OpenMP : actif
Nombre de processus utilis?s : 4
Librairie HDF5 : non disponible
Librairie MED : non disponible
Librairie MFront : non disponible
Version de la librairie MUMPS : 4.10.0
Version de la librairie PETSc : 3.2.0p7
Version de la librairie SCOTCH : 5.1.10

M?moire limite pour l'ex?cution : 2142.86 Mo
consomm?e par l'initialisation : 570.39 Mo
par les objets du jeu de commandes : 0.00 Mo
reste pour l'allocation dynamique : 1570.92 Mo
Taille limite des fichiers d'?change : 12.00 Go

During the execution of the CdA parallel version, also following warning is displayed:

WARNING: There are more than one active ports on host 'frontep2', but the
default subnet GID prefix was detected on more than one of these
ports. If these ports are connected to different physical IB
networks, this configuration will fail in Open MPI. This version of
Open MPI requires that every physically separate IB subnet that is
used between connected MPI processes must have different subnet ID
values.

mpirun has exited due to process rank 0 with PID 27986 on
node frontep2 exiting improperly. There are three reasons this could occur:

1. this process did not call "init" before exiting, but others in
the job did. This can cause a job to hang indefinitely while it waits
for all processes to call "init". By rule, if one process calls "init",
then ALL processes must call "init" prior to termination.

2. this process called "init", but exited without calling "finalize".
By rule, all processes that call "init" MUST call "finalize" prior to
exiting or it will be considered an "abnormal termination"

3. this process called "MPI_Abort" or "orte_abort" and the mca parameter
orte_create_session_dirs is set to false. In this case, the run-time cannot
detect that the abort call was an abnormal termination. Hence, the only
error message you will receive is this one.

This may have caused other processes in the application to be
terminated by signals sent by mpirun (as reported here).

The same procedure for installation of CdA sequential and parallel version has been previously successful applied on other RedHat computers without problems.

Is the problem with HDF5 and MED - marked as not present - of the CdA parallel version related to the Open MPI problem or are this two independent problems ? How should I proceed ?

Greetings
Viktor

Last edited by Viktor.Vlaski (2016-04-27 06:56:22)

mathieu.courtois · 2016-04-12 22:10:03

Hello,

Did you progress on this problem?

If not, join the output of "waf configure".

MC

Viktor.Vlaski · 2016-04-25 14:14:46

Hello Mathieu,

unfortunately, I did not make any progress ...

I am attaching the logfile. Out of attachment config.log it seems that it does not find the python2.7 library ():
err: /usr/bin/ld: cannot find -lpython2.7
collect2: error: ld returned 1 exit status

On the other side, the parallel CdA version is compiled without any error message, I can even start it but it crashes then with the error message, shown in my first post.

A short additional explanation. I am trying to compile CdA on a cluster on which pyton2.6 is installed. Other python version can not be installed due to system stability for other software present. In order to use python2.7 for CdA, I exported through LD_LIBRARY_PATH all paths where python2.7 libraries of Salome_Meca (already installed on the cluster) are present. Also through PATH variable the binary path is adopted to use python2.7 of Salome_Meca. This procedure works perfect for the sequential version of CdA. As explained above, also the parallel version is compiled without any error, but when the parallel version is executed, it does not find hdf5 and med libraries and crashes.

Greetings
Viktor

Attachments:
config.log, Size: 341.97 KiB, Downloads: 452

mathieu.courtois · 2016-04-25 15:58:11

Hello,

Error messages in config.log are normal. This is its job to try different possibilities to find those it works.
That's why I asked the output (not its log file) of 'waf configure'.

The error message means that mpirun does not start properly.
IB=infiniband ? probably you need a specific environment... Perhaps this environment is required to link Code_Aster against the IB libs...?

MC

Viktor.Vlaski · 2016-04-27 06:55:26

Hello,

the reason were compile options for OpenMPI:
after adding --disable-shared --enable-static to OpenMPI configure (proplem with libmpi plugin https://www.open-mpi.org/faq/?category=troubleshooting) it works ...

Regards
Viktor

Last edited by Viktor.Vlaski (2016-04-27 06:58:51)

#1 2016-03-22 07:45:56

[SOLVED]CdA Parallel Version 12.5 does not find HDF5 and MED libraries

#2 2016-04-12 22:10:03

Re: [SOLVED]CdA Parallel Version 12.5 does not find HDF5 and MED libraries

#3 2016-04-25 14:14:46

Re: [SOLVED]CdA Parallel Version 12.5 does not find HDF5 and MED libraries

#4 2016-04-25 15:58:11

Re: [SOLVED]CdA Parallel Version 12.5 does not find HDF5 and MED libraries

#5 2016-04-27 06:55:26

Re: [SOLVED]CdA Parallel Version 12.5 does not find HDF5 and MED libraries

Board footer