WRF quitting every three hours

[MichaelIgb]

Hey, I've been trying to run high-resolution WRF for the TC case (so there is also moving nests), and I've noticed that despite me cleaning and recompiling both WRF and WPS and regenerating new BC files for the model, it's still not running correctly.

Essentially it has a SIGTERM error roughly every three hours, if it doesn't have one trying to use a restart file, and I have no idea why it's doing this. Could it be something in my namelist that's really obscure but sure to be causing problems? Thank you so much in advance, I've attached rsl output files, my namelist, and my slurm job file.

 

[Ming Chen]

Please confirm that your namelist and rsl files are correct. I ask tyhis becaus ein your rsl file, it shows that

Domain # 1: dx = 30000.000 m

In your namelist.input,

dx = 2250, 750

 

[MichaelIgb]

Please confirm that your namelist and rsl files are correct. I ask tyhis becaus ein your rsl file, it shows that

Domain # 1: dx = 30000.000 m

In your namelist.input,

dx = 2250, 750

These are the right files, i just looked at them again and resolutions match. I'll reattach

 

[Ming Chen]

There is no other error message except " forrtl: error (78): process killed (SIGTERM)" at the end of your rsl file. This error often indicates that something is wrong in your physics. Looking at your namelist.input, I found that
(1) start_hour = 03, 03
start_second = 00, 01
Can you set start_second = 00, 00, then try again?
(2) Please turn off the adaptive time step option, --- sometimes this option can cause problems.
(3) Your model top is 50 hPa and you run this case with 201 vertical levels, which implies that the vertical resolution is high. This might affect the stability of the model integration. Can you reduce the vertical levels?

If the case still failed with above changes, please recompile WRF in debug mode (i.e., ./clean -a and ./configure -D), then rerun this case. With debug mode, you will get information when and where the model crashes first. This will provide some hints to figure out what is wrong.

 

[MichaelIgb]

There is no other error message except " forrtl: error (78): process killed (SIGTERM)" at the end of your rsl file. This error often indicates that something is wrong in your physics. Looking at your namelist.input, I found that
(1) start_hour = 03, 03
start_second = 00, 01
Can you set start_second = 00, 00, then try again?
(2) Please turn off the adaptive time step option, --- sometimes this option can cause problems.
(3) Your model top is 50 hPa and you run this case with 201 vertical levels, which implies that the vertical resolution is high. This might affect the stability of the model integration. Can you reduce the vertical levels?

If the case still failed with above changes, please recompile WRF in debug mode (i.e., ./clean -a and ./configure -D), then rerun this case. With debug mode, you will get information when and where the model crashes first. This will provide some hints to figure out what is wrong.

I fixed 2 and 3 then completely restarted, it now works perfectly fine. Thank you ! I ran with 100 vertical levels FWIW