gmake error when building CAM5 on yellowstone

[tfan@colorado_edu]

Dear all,I met a problem when I run CAM5_3 on yellowstone. The build script looks like this:#! /bin/tcsh -f

source /glade/apps/opt/lmod/lmod/init/tcsh

module purge
module load ncarenv/1.0
module load ncarbinlibs/1.1
module load intel/12.1.5
module load ncarcompilers/1.0
module load netcdf/4.3.0
module load idl/8.2.3
module load ncl/6.2.0
module load nco/4.4.2
module load esmf
module load ncview/2.1.1

.... $cfgdir/configure
      -cc mpicc -fc mpif90 -fc_type intel
      -phys cam5
      -chem trop_mam3
      -spmd
      -nosmp
      -dyn fv
      -res 0.9x1.25 -cice_bsizex 3 -cice_bsizey 192 -cice_maxblocks 32 -cice_decomptype cartesian
      || echo "configure failed" && exit 1  /usr/bin/gmake -j8 >&! MAKE.out      || echo "CAM build failed: see $blddir/MAKE.out" && exit 1echo "Create the namelist..."
echo " "
cd $rundir                      || echo "cd $rundir failed" && exit 1
$cfgdir/build-namelist -verbose -case $case -runtype 'startup' -d $rundir 
  -config $blddir/config_cache.xml
  -namelist "&camexp  stop_n=12, stop_option='nmonths'
  npr_yz = 60,4,4,60
 / "  || echo "build-namelist failed" && exit 1

 ....
When running the build script, when it comes to make, the terminal do not respond to any input including "ctr-C". The MAKE.out looks like that all the programs have been compiled sucessfully, all the .o .mod file look good. But it never comes to the build-namelist step. I tried running configure in the commond line sucessfully. I asked someone to run the same code and the same script from another terminal in another place, the build scripts finishes sucessfully in 5 mins.
Does anyone know how to solve this problem? Cheers,Tianyi  

 

[jedwards]

try using the CESM build instead?

 

[eaton]

Did the cam executable file get built?Sometimes you need to search the Make.out file for the error message.  It's not always visible right at the end of the file when doing a parallel build. 

 

[santos]

If it's just today, switch login nodes? Maybe yours isn't talking very well to either the filesystem or license server.

 

[tfan@colorado_edu]

Hi Eaton,Yes, the cam excutable got built . Does it usually mean make susceed? I see several warning messages in the MAKE.out file, but not sure if they matter. I tried build-namelist after the make from the command line, it seems to work. All the desired namelists are generated.
Attached is the MAKE.out file.  

 

[tfan@colorado_edu]

 Hi Sean,Thanks for pointing this out. But it is not just today. I tried login2 and login6, the same issue happened.

 

[santos]

You should try different module definitions. For instance, intel/13.1.2 probably works, and intel/15.0.0 will work for a very new CAM tag, but intel/12.1.5 will not.The CESM scripts should have the correct modules for yellowstone, so if you take Jim's suggestion it will probably work. Or you can copy them by looking in scripts/ccsm_utils/Machines/env_mach_specific.yellowstone

 

[tfan@colorado_edu]

Dear Jim, Eaton, and Sean,It turns to be a terminal problem. I use Xterm in Mac which disconnects from the remote server if no input for a while. It must get disconnected before build-namelist is done. That explains why I see cam excutables but no *_in namelists.  It is totally an unexpected issue. When I switch to iterm (http://iterm2.com), it finishes building.Thank you very much for your comments and suggestions. I learned a lot. Thank you!
Cheers,Tianyi  

 

[jedwards]

 Glad you figured it out.   If you add ServerAliveInterval 300to the .ssh/cronfig file on your client, I think it will solve your xterm problem.

 

[eaton]

I am able to run cesm1_2_2 with the intel/12.1.5 compiler, at least for F compsets or standalone CAM.  I know that current development code doesn't work with 12.1.5, but it seems that cesm1_2_2 is OK.