Mesh import error on cluster
Posted: Fri Feb 28, 2014 11:22 am
Hello,
I am now running code_saturne 3.1.0 on a PBS system cluster. The GUI is not supported so I am having to run things via the command line. This is the procedure I have been following (provided by the Cluster Wiki) I have emboldened my comments below to try and make it easier to read:
module load xe-code-saturne/3.1.0
1 Create a study with a case
2 Copy the mesh in MESH
3 Copy the usersubroutines in SRC
4 Go to DATA and copy REFERENCE/cs_user_scripts.py in DATA
5 Edit cs_user_scripts.py and:-
- At L.128, change domain.mesh_input from None to the mesh path, using quotes
- At L.251, set the number of processors for the simulation
- At L.297, type mpi_env.mpiexec = 'aprun '
- At L.311, type mpi_env.mpiexec_n = ' -n '
6 Save and close cs_user_scripts.py
7 At the case level, type code_saturne run --initialize
8 In RESU should be created a directory which name is based on the date code_saturne run --initialize was run. Go to this directory, edit run_solver.sh and adapt the following to your case (using PBS instructions):
#-------------------------------------------------------------------------------
#
#PBS -A budget
#PBS -N casename
#PBS -l mppwidth=nprocs
#PBS -l mppnppn=ncores
#PBS -l walltime=hours:minutes:00
#PBS -j oe
#
9 The job is then submitted typing qsub run_solver
However, I get an error saying that c_s cannot read my mesh:
/usr/local/packages/nag/saturne/SATURNE_3.1.0_GNU_XE6/SATURNE_3.1.0/SATURNE_3.1.0/saturne-3.1.0/src/base/cs_io.c:702: Fatal error.
Error reading file: "mesh_input".
File format is not the correct version.
The first 64 bytes expected contain:
"Code_Saturne I/O, BE, R0"
The first 64 bytes read contain:
"HDF
"
From reading past forum posts Yvan said that this was cause by hardcoding the mesh path in cs_userscripts. However if I leave that function empty then it wont compile - saying mesh cannot be found.
Any ideas?
Thanks in advance for your help!
Susan
I am now running code_saturne 3.1.0 on a PBS system cluster. The GUI is not supported so I am having to run things via the command line. This is the procedure I have been following (provided by the Cluster Wiki) I have emboldened my comments below to try and make it easier to read:
module load xe-code-saturne/3.1.0
1 Create a study with a case
2 Copy the mesh in MESH
3 Copy the usersubroutines in SRC
4 Go to DATA and copy REFERENCE/cs_user_scripts.py in DATA
5 Edit cs_user_scripts.py and:-
- At L.128, change domain.mesh_input from None to the mesh path, using quotes
- At L.251, set the number of processors for the simulation
- At L.297, type mpi_env.mpiexec = 'aprun '
- At L.311, type mpi_env.mpiexec_n = ' -n '
6 Save and close cs_user_scripts.py
7 At the case level, type code_saturne run --initialize
8 In RESU should be created a directory which name is based on the date code_saturne run --initialize was run. Go to this directory, edit run_solver.sh and adapt the following to your case (using PBS instructions):
#-------------------------------------------------------------------------------
#
#PBS -A budget
#PBS -N casename
#PBS -l mppwidth=nprocs
#PBS -l mppnppn=ncores
#PBS -l walltime=hours:minutes:00
#PBS -j oe
#
9 The job is then submitted typing qsub run_solver
However, I get an error saying that c_s cannot read my mesh:
/usr/local/packages/nag/saturne/SATURNE_3.1.0_GNU_XE6/SATURNE_3.1.0/SATURNE_3.1.0/saturne-3.1.0/src/base/cs_io.c:702: Fatal error.
Error reading file: "mesh_input".
File format is not the correct version.
The first 64 bytes expected contain:
"Code_Saturne I/O, BE, R0"
The first 64 bytes read contain:
"HDF
"
From reading past forum posts Yvan said that this was cause by hardcoding the mesh path in cs_userscripts. However if I leave that function empty then it wont compile - saying mesh cannot be found.
Any ideas?
Thanks in advance for your help!
Susan