Cannot define variable, h when creating restart file

[d.kobashi]

Hi,

I am having a strange error while running ROMS.
When I turn off USE_DEBUG in coawst.bash, the compile goes well, but I get the following error while ROMS tries to create a restart file.

Code: Select all

   STEP   Day HH:MM:SS  KINETIC_ENRG   POTEN_ENRG    TOTAL_ENRG    NET_VOLUME
          C => (i,j,k)       Cu            Cv            Cw         Max Speed

      0 14610 00:00:00  8.331972E-03  7.785941E+03  7.785949E+03  1.202318E+14
          (547,028,30)  4.212970E-02  5.095988E-02  0.000000E+00  1.303715E+00
      DEF_STATION - creating stations file, Grid 01: /scratch/user/d.kobashi/projects/hindcasts/projects/txla2/outputs/2010/ocean_sta.nc

DEF_VAR - unable to define variable: h
           in NetCDF file: /scratch/user/d.kobashi/projects/hindcasts/projects/txla2/outputs/2010/ocean_rst.nc

 Elapsed CPU time (seconds):

 Node   #  0 CPU:       5.184

 ROMS/TOMS - Output NetCDF summary for Grid 01:
             number of time records written in STATION file = 00000001

 Analytical header files used:

     /home/d.kobashi/forecasts/src/Functionals/ana_btflux.h
     /home/d.kobashi/forecasts/src/Functionals/ana_sponge.h
     /home/d.kobashi/forecasts/src/Functionals/ana_nudgcoef.h
     /home/d.kobashi/forecasts/src/Functionals/ana_stflux.h

 ROMS/TOMS - Output error ............ exit_flag:   3

 Node   #  1 CPU:      68.800
 Node   #  2 CPU:      68.799
 Node   #  3 CPU:      68.799

 ERROR: Abnormal termination: NetCDF OUTPUT.
 REASON: NetCDF: Invalid dimension ID or name    

The error message indicates that the model cannot define h (bathymetry) while trying to create a restart file due to invalid dimension ID or name. Thus, the model failed to create restart file.

However, when I turn on USE_DEBUG, the model creates restart file without any errors.
I have never had this problem before.

It is strange because the error is nothing to do with computation (e.g. floating point error etc.).

I assume the difference between USE_DEBUG off and USE_DEBUG on is the optimization of the compiler.
I use intel fortran/intel mpi to run my model.

My FFLAGS without USE_DEBUG is

Code: Select all

FFLAGS += -ip -O2 

My FFLAGS with USE_DEBUG is

Code: Select all

FFLAGS += -g -check uninit -ftrapuv -traceback

The following line is used by both

Code: Select all

FFLAGS := -heap-arrays -fp-model fast -mt_mpi

The grid parameters are defined in def_info.F and it is nothing to do with debug mode. def_info.f90 in Build directory has the lines defining bathymetry, h, but the model execution did not pass def_var resulting in the error.

Code: Select all

1245 !
1246 !  Bathymetry.
1247 !
1248         Vinfo( 1)='h'
1249         Vinfo( 2)='bathymetry at RHO-points'
1250         Vinfo( 3)='meter'
1251         Vinfo(14)='bath, scalar'
1252         Vinfo(22)='coordinates'
1253         Aval(5)=REAL(r2dvar,r8)
1254         IF (ncid.eq.STA(ng)%ncid) THEN
1255           status=def_var(ng, model, ncid, varid, NF_TYPE,               &
1256      &                   1, (/stadim/), Aval, Vinfo, ncname)
1257           IF (exit_flag.ne.NoError) RETURN
1258         ELSE
1259           status=def_var(ng, model, ncid, varid, NF_TYPE,               &
1260      &                   2, t2dgrd, Aval, Vinfo, ncname)
1261           IF (exit_flag.ne.NoError) RETURN
1262         END IF

What's the difference between USE_DEBUG=on and off aside from Fortran optimization in ROMS?

I've recently added additional variables to history files and that might be something to do with the error. Did the addition mess up netCDF id or name? The addition is to do with history file and is nothing to do with restart file, though.

I can still run the model with USE_DEBUG, but this issue is bugging me.
The model I use is COAWST and revision is 960.

Any suggestions and/or comments would be appreciated.

Thanks in advance.

-DJ

[jcwarner]

not sure what is going on here. Can you try it without the changes that you added?
-john