[Wrf-users] Intermittent WRF failures

[Richard Carpenter]

We have an older system with PGI Fortran and SGE that we are successfully running version 3.3.1 on, but we are having problems upgrading to 3.5.1 or 3.7.1. Often, but not always, WRF will fail right before time stepping begins. We have tried to isolate this but so far we have been unsuccessful. We have tried turning off FDDA, grid nesting, history dumps, physics and dynamics options (including radiation), etc. We have tried varying the number of nodes, node list, number of processors used, and use of an I/O node.  We have also run it with consistent success on other systems. I’ve provided some of the error output below. If anyone has any ideas, please let me know. Thanks!

Tail of rsl.error.0000 (debug=999):
WRF TILE   1 IS      1 IE    182 JS      1 JE      4
WRF NUMBER OF TILES =   1
d01 2016-03-01_14:00:00  grid spacing, dt, time_step_sound=    12000.00        100.0000                6
d01 2016-03-01_14:00:00  call rk_step_prep
d01 2016-03-01_14:00:00 calling inc/HALO_EM_A_inline.inc
d01 2016-03-01_14:00:00 calling inc/PERIOD_BDY_EM_A_inline.inc
d01 2016-03-01_14:00:00  call rk_phys_bc_dry_1
d01 2016-03-01_14:00:00  call init_zero_tendency

Backtrace file wrf.exe.80s-7214,compute-5-0.local.btr:
wrf.exe:7214 terminated with signal 11 at PC=3b07e797e0 SP=7fff4a7ec9c8.  Backtrace:
/lib64/libc.so.6(strlen+0x30)[0x3b07e797e0]
/lib64/libc.so.6(getenv+0x25)[0x3b07e32a55]
/share/nwpdata3/apps/WRF_Model/v3_7_1/ncar/WRFV3/main/wrf.exe(pgf90_get_env_var+0x72)[0x33311c2]

Richard Carpenter
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.ucar.edu/pipermail/wrf-users/attachments/20160301/0e288f48/attachment.html