Error in boundary condition specification with real.exe

[TheodorB]

I'm trying to run a real 1 domain simulation over the eastern Mediterranean. I'm stuck in the real.exe part because of the following error
from rsl.error.0000:
########################################################################################
01 2017-10-19_00:00:00 Timing for input 0 s.
d01 2017-10-19_00:00:00 flag_soil_layers read from met_em
d01 2017-10-19_00:00:00 Turning off use of MAX WIND level data in v
d01 2017-10-19_00:00:00 Turning off use of TROPOPAUSE level data in
Max map factor in domain 1 = 1.15. Scale the dt in the model accor
*** Error in boundary condition specification
boundary conditions at xs 0
boundary conditions at xe 0
boundary conditions at ys 0
boundary conditions at ye 0
boundary conditions logicals are
periodic_x F
periodic_y F
symmetric_xs F
symmetric_xe F
symmetric_ys F
symmetric_ye F
open_xs F
open_xe F
open_ys F
open_ye F
polar F
nested F
specified F
-------------- FATAL CALLED ---------------
FATAL CALLED FROM FILE: <stdin> LINE: 165
*** Error in boundary condition specification
-------------------------------------------
application called MPI_Abort(MPI_COMM_WORLD, 1) - process 1
#####################################################################################

my namelist.input has this:
#################################
specified = .true., .false., .false.,
nested = .false., .true., .true.,
#################################

defined so that is not the problem.

Thanks,
Theodor

[kwthomas]

Theodor.,..

Are your boundary files >2gb and you are using a version of WRF earlier than 3.9? If so
you need to build the software with the environmental variable WRFIO_NCD_LARGE_FILE_SUPPORT set to 1. WRF people finally made this a default
in WRF 3..9. In earlier versions, without the environmental variable set, WRF creates 32-bit NETCDF files. When the output file is >2gb, index locations will go negative, meaning your NETCDF file is garbage. Setting the environme variable creates 64-bit files.

If this isn't the problem, here is something else to try. Untar the WRF software into a new
directory tree. Copy the "config.wrf" file from your build to the new directory. Edit it and look
for the FCOPTIM line. Change -O3 to -O0. This will build the software will no Fortran optimization. This is assuming the Intel compiler. Try running the executable and see what
happens.

I'm having problems on STAMPEDE2 (TACC) on the MIC nodes with any optmization levels. The problem is different than yours, but the cause may be the same, an optmization issue. I'm still trying to figure out what is happening. I can reproduce the problem with
WRF 3.7.1 (CAPS version), WRF 3.8.1) (CAPS version) and WRF 3.9.1.1 (out of the box).

[TheodorB]

Thanks for the quick reply!

I have the latest WRF 3.9.1.1 compiled with gfortran. I'm trying to run just a simple 24h run with GFS data.

Theodor

[TheodorB]

And I just checked to have
export WRFIO_NCD_LARGE_FILE_SUPPORT=1

defined in my .bashrc file

Theodor

[TheodorB]

Well, all is OK now. I had a mistype in my namelist.input that made my bdy_control part unreadable.

SOLVED