Segmentation error/Cannot place all ranks on node list

[bashaman]

Hi,

I have previously successfully run my WRF model with the current settings I have, only with a smaller set of nested grids. In expanding the grids to cover a larger area, I am now running into new errors. I have adhered to "Choosing an Appropriate Number of Processors" as described in FAQ. When I do more cores, I get an error in the wrf.o* file saying "Cannot place all ranks on node list" and when I do fewer cores, I run into a segmentation error.

I have attached files corresponding to the segmentation error occurrence. Please let me know if there's anything else I can attach that would be helpful. Thanks in advance for any help.

- Andrew

 

[kwerner]

Hi Andrew,
I see that you're running this on Derecho, meaning you have a possibility of using 128 processors per node. As a test, can you try to use 1280 processors - so set

#PBS -l select=10:ncpus=128:mpiprocs=128

and see if that makes any difference. Using this many processors allows the distribution to be an even 32x32 processors, and sometimes have a "more squared" decomposition can help, but not always. If not, do you mind pointing me to the directory where you're running this, so I can take a closer look?

 

[bashaman]

Hi, I realized that I had max_dom = 2 instead of max_dom =3 in my namelist.input. It is good to know that Derecho has 128 ppn.

Now I am running into a new error saying "real_em.F: Could not find the parent domain", which appears to be an unresolved issue being discussed in another post (real_em.F: Could not find the parent domain). Any insights into that problem would be helpful.

Thanks!

 

[kwerner]

It's an interesting issue. If you're okay sharing the path(s) to your running directories on derecho, I can try to run my own test, using your namelist and files to try to recreate the error - that would help me to be able to troubleshoot easier.

 

[bashaman]

Here are my running directories:
/glade/work/bashaman/ms_thesis/wrfv4.6.0/wpsv4.6.0
/glade/work/bashaman/ms_thesis/wrfv4.6.0/run

 

[kwerner]

Thanks for sharing those. I believe the issue is that you are missing the parameter "grid_id" in your namelist.input file. Add this to the &domains section and try to run real again:

grid_id = 1, 2, 3