Program received signal SIGSEGV: Segmentation fault - invalid memory reference.

[kermarrec]

I am facing a segmentation problem (although I extended to unlimited) after 1 hour of simulation. It is a real case with LES and 5 domains
Program received signal SIGSEGV: Segmentation fault - invalid memory reference.

Backtrace for this error:
#0 0x7fed117fbd11 in ???
#1 0x7fed117faee5 in ???
#2 0x7fed112e708f in ???
at /build/glibc-SzIz7B/glibc-2.31/signal/../sysdeps/unix/sysv/linux/x86_64/sigaction.c:0
#3 0x5608f5537d3f in ???
#4 0x5608f553d304 in ???
#5 0x5608f5541382 in ???
#6 0x5608f4d1f995 in ???
#7 0x5608f4548525 in ???
#8 0x5608f3f8f498 in ???
#9 0x5608f3e248f3 in ???
#10 0x5608f2e666ad in ???
#11 0x5608f2e66d1a in ???
#12 0x5608f2e66d1a in ???
#13 0x5608f2ded667 in ???
#14 0x5608f2ded09e in ???
#15 0x7fed112c8082 in __libc_start_main
at ../csu/libc-start.c:308
#16 0x5608f2ded0dd in ???
#17 0xffffffffffffffff in ???


May it be that the problem comes from my setting? I attach the namelist and the error file. Thanks for your help.

 

[kwerner]

According to your rsl* file, you are only using a single processor for this simulation.

 Ntasks in X            1 , ntasks in Y            1

But you should be using more. Take a look at Choosing an Appropriate Number of Processors.

 

[kermarrec]

I have changed the input file as attached by letting
nproc_x =2,
nproc_y =2,
/
and
&namelist_quilt
nio_tasks_per_group = 1,
nio_groups = 2,

but now, I face this error, which occurs by real.exe
taskid: 0 hostname: kornog
module_io_quilt_old.F 2931 T
-------------- FATAL CALLED ---------------
FATAL CALLED FROM FILE: <stdin> LINE: 5874
check comm_start, nest_pes_x, nest_pes_y settings in namelist for comm 1
-------------------------------------------

if I do not add nproc_x and _y and now get the following error after running real.exe successfully for wrf.exe:


taskid: 0 hostname: kornog
module_io_quilt_old.F 2931 F
MPASPECT: UNABLE TO GENERATE PROCESSOR MESH. STOPPING.
PROCMIN_M 1
PROCMIN_N 1
P 0
MINM 1
MINN 0
-------------- FATAL CALLED ---------------
FATAL CALLED FROM FILE: <stdin> LINE: 124
module_dm: mpaspect
-------------------------------------------

I don't know what suddently happened as nothing works now. There is probably a mistake somewhere? Should I recompile wrf? where does the problem come from?
thanks for your help!
best

 

[kermarrec]

I managed to come further but it still does not work. I recompiled the wrf and used the command mprun -np 4 ./wrf.exe instead of nproc_x which did not work.
Now I am still having a segmentation fault error. I attach the rsl and the namelist.
I ran grep cfl rsl* and did not get message. However I tried several things: w_damping=1, smooth_cg_topo =.true, decrease the time step.
Maybe you have some idea how I can come further. Many thanks in advance, I would appreciate any help to be able to lauch this simulation

 

[kwerner]

Can you let me know the resolution of your input data (the data you use in the ungrib process)?

 

[kermarrec]

I used the default setting of WPS (for HGT_M default:topo_gmted2010_30s and for LANDUSEF default:modis_landuse_20class_30s_with_lakes/

Is it not enough for the 1000 m domain (because it is at that point that I get this error)?

 

[kermarrec]

I changed to modis_15s_lake for LANDUSEF and topo_gmted2010_30s for HGT_M and unfortunately, it still did not work and stopped at the same place. I attach the error file.

 

[kwerner]

Hi,
I was referring to the resolution of the meteorological input data that you use for ungrib (e.g., ERA5 or GFS, etc) - not the static fields during geogrid. Can you let me know the resolution of the meteorological input? Thanks!

 

[kermarrec]

I used the
NCEP GDAS/FNL 0.25 Degree Global Tropospheric Analyses and Forecast Grids
should I rather use ERA5?

 

[kwerner]

Apologies for the delay. The data you're using is fine, but 0.25 degree data has a grid distance of ~28 km, and your parent domain (d01) uses a grid distance of 3 km. This is a 9:1 ratio. We recommend trying to stay in a ratio range no larger than about 5:1. This could be the reason for your issue. You may need an outer domain around the 3km domain (perhaps something like 9km) so there's not so much of a difference between the resolution of the input data and your parent domain.

 

[mzapatahe]

Hello

I am getting the same error, you could say me how you fixed it ?

 

[kermarrec]

Hi, there was a problem with my meteodata that was not going with the domain. Sometimes it is also the timestep which needs to be adapted. I would check that. good luck gael.