NEED HELP!!

[roknavy]

I am a new ROMS user. I just tried to follow roms set up video instruction but below error came out when I excuted ./oceanM < ROMS/External/ocean_upwelling.in. after running make.
Would you tell me what is my problem and how to solve it?

Node # 0 (pid= 17156) is active.

Model Input Parameters: ROMS/TOMS version 3.0
Friday - September 12, 2008 - 12:47:17 AM
-----------------------------------------------------------------------------

INP_PAR - Unable to open ROMS/TOMS input script file.
In distributed-memory applications, the input
script file is processed in parallel. The Unix
routine GETARG is used to get script file name.
For example, in MPI applications make sure that
command line is something like:

mpirun -np 4 ocean ocean.in

and not

mpirun -np 4 ocean < ocean.in

[arango]

The fact that you created the oceanM executable implies that you compiled with the parallel MPI library. In these cases you need to run with mpirun. There is plenty of information about this in wikiROMS, you need to do your homework and read.

[luis.delcastillo]

Mr arango i edit the build.sh script and change the setenv HDF5_LIBDIR /usr/local/hdf5/hdf5/lib where is the location of my libhdf5_hl.so.0. When i run ./build.sh
everything compile good so depending if i uncoment the USE_MPI the compilation make the file oceanM when i try to run it with mpirun -np 5 oceanM ocean_upwelling.in they send me an error like unable to locate libhdf5_hl.so.0.

When i coment the USE_MPI the compilation make the oceanS and run very well.
also i just figure it out that when i use ./oceanM ocean_upwelling.in alone without mpirun script also compile good.

do you have any idea why this is happening

[kate]

For shared libraries, you need to tell it the location at compile time and also at run time. You've got the compile time one working fine. The run time is through the LD_LIBRARY_PATH environment variable. This will have to be set in your job script for the parallel job, if that's how you run parallel jobs.

[luis.delcastillo]

hi kate in fact in my .tcshrc i have

alias ls 'ls --color'
setenv NCARG_ROOT /usr/local
setenv PGI /usr/local/pgi
setenv PATH "$NCARG_ROOT/lib:$NCARG_ROOT/bin:$NCARG_ROOT/include:/usr/local/pgi/default:/usr/local/pgi/default/bin:/usr/local/pgi/default/include:/usr/local/pgi/default/lib:/usr/local/pgi/default/jre/bin:/usr/local/mpich:/usr/local/mpich/bin:/usr/local/mpich/lib:/usr/local/mpich/include:/bin:/sbin:/usr/sbin:$PATH"

setenv LD_LIBRARY_PATH /usr/lib:/usr/lib32:/usr/local/hdf5/hdf5/lib
limit stacksize unlimited
setenv LM_LICENSE_FILE /usr/local/pgi/license.dat
setenv LIB_MPI /usr/local/mpich/lib
setenv INC_MPI /usr/local/mpich/include
setenv P4_RSHCOMMAND ssh
setenv PGI_JAVA /usr/local/pgi/default/jre/bin
setenv NETCDF_LIBDIR /usr/local/lib
setenv NETCDF_INCDIR /usr/local/include
setenv HDF5_LIBDIR /usr/local/hdf5/hdf5/lib

[kate]

Yes, that explains why it runs fine in serial mode. In parallel mode, you are executing the mpirun script to run the job. Check that script. Is it setting LD_LIBRARY_PATH so your parallel job can find the mpich library? You might have to edit it to include your hdf path as well.

[luis.delcastillo]

look my mpich1.2.7p1 is working properly

[ocean@cl2master:~]> mpirun -np 8 /usr/local/mpich/examples/cpi
Process 2 on node2
Process 3 on node3
Process 1 on node1
Process 5 on node5
Process 6 on node6
Process 0 on cl2master
Process 4 on node4
Process 7 on node7
pi is approximately 3.1416009869231249, Error is 0.0000083333333318
wall clock time = 0.011719

[linzhenhua]

The cpi example works because it does not need the HDF library.

[luis.delcastillo]

yes i just realise that i need to compile hdf5 and netcdf4 to use parallel like this viewtopic.php?f=13&t=1181