Welcome, Guest
Username: Password: Remember me
HELPDESK

Here we can describe more what should be posted here

TOPIC: cy43 Forecast crashing due to zero divide

cy43 Forecast crashing due to zero divide 1 year 1 month ago #2486

I ran into an error in the Forecast task. I think it is in the surface part of the forecast which crashes. As it uses EXSEG1.nam namelist.

Notes:
- Latest harmon_cy43h_21 branch is used on cca, together with METCOOP observations. Though it gives the same error with only MARS observations.
- I am using FirstGuess-files from MEPS (Upper Air and Surface), which I transferred to ecfs manually and the Data Assimilation on those files works.
- There seems to be an
" Program received signal SIGFPE: Floating-point exception - erroneous arithmetic operation. " which usually indicates a zero divide somewhere.



traceback:

tid#1 starting drhook traceback, time =1607528759.33
[myproc#45,tid#1,pid#56279]: 4287 MB (maxheap), 32 MB (maxrss), 0 MB (maxstack), walltime = 1607528759.33s
[myproc#45,tid#1,pid#56279]: MASTER
[myproc#45,tid#1,pid#56279]: CNT0
[myproc#45,tid#1,pid#56279]: CNT1
[myproc#45,tid#1,pid#56279]: CNT2
[myproc#45,tid#1,pid#56279]: CNT3
[myproc#45,tid#1,pid#56279]: CNT4
[myproc#45,tid#1,pid#56279]: STEPO
[myproc#45,tid#1,pid#56279]: SCAN2M
[myproc#45,tid#1,pid#56279]: GP_MODEL_STACK
[myproc#45,tid#1,pid#56279]: GP_MODEL
[myproc#45,tid#1,pid#56279]: CPG_DRV
[myproc#45,tid#1,pid#56279]: CPG
[myproc#45,tid#1,pid#56279]: MF_PHYS
[myproc#45,tid#1,pid#56279]: APL_AROME
[myproc#45,tid#1,pid#56279]: ARO_GROUND_PARAM
[myproc#45,tid#1,pid#56279]: COUPLING_SURF_ATM_N
[myproc#45,tid#1,pid#56279]: COUPLING_SURF_ATM_n:TREAT_SURF
[myproc#45,tid#1,pid#56279]: COUPLING_NATURE_N
[myproc#45,tid#1,pid#56279]: COUPLING_ISBA_SVAT_N
[myproc#45,tid#1,pid#56279]: COUPLING_ISBA_OROGRAPHY_N
:[myproc#45,tid#1,pid#56279]: COUPLING_ISBA_CANOPY_N
[myproc#45,tid#1,pid#56279]: COUPLING_ISBA_N
[myproc#45,tid#1,pid#56279]: COUPLING_ISBA_n:TREAT_PATCH
[myproc#45,tid#1,pid#56279]: DIAG_INLINE_ISBA_N
[myproc#45,tid#1,pid#56279]: CLS_TQ
JSETSIG: sl->active = 0
[drhook.c-l1133] signal_harakiri(SIGALRM=14): New handler installed at 0x20aace60; old preserved at (nil)
***Received signal = 8 and ActivatED SIGALRM=14 and calling alarm(10), time =1607528759.90



Maybe anyone ran into a similar problem already and can provide help.

Log file is attached, thanks.

filebin.net/evsnu4704t70gkjn


David/
Last Edit: 1 year 1 month ago by David Schönach.

cy43 Forecast crashing due to zero divide 1 year 1 month ago #2487

The problem was caused by incompatible climate files. Initially they were generated on cca and they did not like the FG files from MEPS, which are generated on a different system. Using the climate files from MEPS solved the issue.
Probably a very Metcoop specific error, but maybe good to know that such zero divided SIGFPE errors in surfex can be caused by incompatible or wrong climate files.

Cheers
The following user(s) said Thank You: Daniel Santos Munoz
Time to create page: 0.073 seconds