metgrid.exe be created but mpirun stuck with no errors or interuptions

[David Jones]

Hello, there,
I am new here.

Follow the topic of 'metgrid.exe not created when compiling WPS',

I have replaced the .F file and WPS compile work, and metgrid.exe be created. But, another problem, I can not 'mpirun -np 4 ./real.exe', it stucked without any error, and 'mpirun -np 1 ./real.exe' work. And I have posted as below:

​

Mpirun stuck with no errors or interuptions​

I got a problem. After running 'mpirun -np 4 ./real.exe', it gets stuck. Four processes can be seen in command 'top', and there are no errors or interruptions, but it just gets stuck. But I tested 'mpirun -np 1 ./real.exe' or just './real.exe' can run normally. And before that I have checked the compilation of WRF and WPS. Have anyone ever encountered this situation？By the way, right now this is only running on one node, let alone running across nodes.

Anyone help? Thanks a lot.

 

[kwerner]

Hi David,
It seems that you have something going on with your environment that isn't allowing you to run with more than one processor, since you have this open issue, as well. I think you need to talk to a systems administrator at your institution to make sure you've got everything set up correctly to run with multiple processors.

That being said, it's not necessary to compile WPS for parallel processing, unless your domains are using 1000s x 1000s of grid spaces (and you're not). Even if you compiled WRF for dmpar, you can compile WPS serially and run each of the executables in the command line (e.g., ./metgrid.exe).