Encountered forrtl while running LES: severe (174): SIGSEGV, segmentation fault occurred

[rubyyuki]

Hello all,

Greetings!

I'm trying a LES simulation in the em_real folder. I did WRFDA assimilation for the initial moments. I am running the simulation on the server and using 120 processes. Due to server math, runtime, and cost issues, I am using a maximum of 120 processes and time_step cannot be set too small.

Before this I set the time_step option to 30s and also had segmentation fault, but at the same time reported cfl error. After I changed the time_step to 20s the cfl error disappeared but the segmentation fault is still there. I have tried what kwerner's post Segmentation Faults - Helpful Information | WRF & MPAS-A Support Forum says, but still can't solve it.

I'm at a loss as to how to resolve this issue and am very much looking forward to your help.

Attached are my namelist.input and rsl.error

Kind regards.

 

[kwerner]

Hi,
I'd like to apologize for the very long delay in response while we've had to tend to other obligations. Thank you for your patience.

I have a few thoughts:
1) You probably don't need to be using 120 processors for a domain the size of your parent domain, but you likely do need at least that many for d02 and d03. See Choosing an Appropriate Number of Processors.
2) I see that your parent domain is using a 25-km grid-spacing. What is the grid-spacing of the input meteorological data you use? If the ratio between the input data and d02 isn't more than about a 5:1 ratio, you may not need the parent domain. If you'd like to try that, you'll need to re-run geogrid and metgrid again, as well.

 

[rubyyuki]

Hi,
I'd like to apologize for the very long delay in response while we've had to tend to other obligations. Thank you for your patience.

I have a few thoughts:
1) You probably don't need to be using 120 processors for a domain the size of your parent domain, but you likely do need at least that many for d02 and d03. See Choosing an Appropriate Number of Processors.
2) I see that your parent domain is using a 25-km grid-spacing. What is the grid-spacing of the input meteorological data you use? If the ratio between the input data and d02 isn't more than about a 5:1 ratio, you may not need the parent domain. If you'd like to try that, you'll need to re-run geogrid and metgrid again, as well.

Hello, kwerner:
Thank you for your response. My simulation utilizes FNL data, so I think the d01 region is necessary for this process. Subsequently, I attempted to run the simulation with a reduced number of processes, but unfortunately, I encountered segmentation faults again. Apart from these segmentation faults, no other errors were reported.

 

[kwerner]

My simulation utilizes FNL data, so I think the d01 region is necessary for this process.

I assume that you mean the GFS FNL one-degree data? If so, you can access higher resolution input data from GFS (0.25 degree data). When higher-resolution input is available, it's recommended to more likely produce realistic results.

If you are set on using the 1 degree FNL data, can you try running a single domain? If that works, try 2, then 3 - to see if any particular domain seems to be causing the issue. And then please package all of your rsl* files into a single *.tar file and attach that. Thanks!