forrtl: error (78): process killed (SIGTERM)

[ivi013]

Hey,

I'm trying to run WRF for one day with a nested area. When running it works fore some time, and I get the outputfiles for the first time step for both domains. But then it crashes. For the error files, it says "forrtl: error (78): process killed (SIGTERM)" (see attached files), and the slurm file says "srun: error: c9-5: task 91: Exited with exit code 174".

The namelist,slurm, rsl.error and rsl.out are attached. Anybody that has an idea of what is wrong?

 

[kwerner]

Hi,
I see several CFL errors in your rsl.error* files. This indicates that the model has become unstable - typically due to strong updrafts and/or complex terrain. When this happens, we typically recommend decreasing your time_step. I see you are already using a decreased step of 4xDX. You can try to go lower than that, and there are some additional suggestions mentioned here.

However, I have a few more suggestions to try first. Your domain is not really set up in an ideal way. A few suggestions, based on what I see:
1) The size of your d01 is MUCH larger than d02. We typically recommend the nest to be about 1/3 of the outer domain. This may not be a problem, as it's often a problem the other way around, but you may want to consider making your nest a bit larger.
2) You are using a 2:1 ratio with feedback turned on. It's not good practice to use an even ratio when using feedback. You can find additional information about that in this presentation. I would suggest using a 3:1 ratio, instead. You could make your outer domain 3 km, and still keep the 1 km nest, but again, you'd likely want to modify the size of those domains.

Take a look at our best practices page that discusses the WPS namelist and suggestions for domain setup (for future reference, there is also a page for the namelist.input file). This will help you to set-up domains that are reasonable.