Error while running init_atmosphere_atmosphere for lbc conditions in regional domain

[518er2005]

Hello everyone, hope everyone is doing good in these hard times. I have encountered a problem when running init_atmosphere_atmosphere for lbc conditions in the regional domain. The error that I am getting is attached as a screenshot. Also the namelist.init_atmosphere and the stream.init_atmosphere is attached. If anyone can help me in getting over this error it will be much appreciated.

With regards,
Debashis

 

[mgduda]

The error message from the screenshot is

execvp error on file init_atmosphere_model (No such file or directory)

Is the init_atmosphere_model file in your directory a symbolic link, and if so, does it point to an init_atmosphere_model executable? You can check this with

ls -lL init_atmosphere_model

In future, it would be helpful to include error messages and directory listings as text in your post and not as as screenshots -- screenshots aren't searchable on the forum, and anyone who would like to quote the error message must type in the message from the screenshot. It's much better if you can copy and paste the text from your terminal into your forum post and quote it for readability.

 

[518er2005]

hello, thanks for the reply. Actually, I did check for the file and its there. What I did is I reinstalled again the MPAS model and now it's working fine. But the problem which I am facing is that after I use metis to create a partition file for 4 MPI tasks I successfully create the file and also further the initial conditions are created, but when I run " mpiexec -n 4 init_atmosphere_model" for creating lbc conditions I get the error as shown in the screenshot. Furthermore when I run with 1 mpi task using info.4 file as just info and running as " mpiexec -n 4 init_atmosphere_model" i don't encounter any error. The error occurs only after using metis. So if it is clarified whether I am doing everything correctly or is my process wrong it will be much appreciated. By the way, I installed metis through the command line "yum install metis".

And ya I will keep in mind, not to share as screenshots.
Anyways thanks.

Cheers,
Debashis

 

[mgduda]

Apologies if I'm misunderstanding the issue, but in the screenshot you originally posted, the only error that I can see is "execvp error ...".

I did check your namelist, and I noticed that your specification of the root of the partition file name was

&decomposition
    config_block_decomp_file_prefix = 'mediterranean.graph.info.part'
/

I think you may need to add a '.' at the end, since MPAS attempts to append the number of MPI tasks to the string specified by 'config_block_decomp_file_prefix' to build a valid file name. Could you try the following?

&decomposition
    config_block_decomp_file_prefix = 'mediterranean.graph.info.part.'
/

In any case, thanks for noting that there is a YUM package for Metis -- I wasn't aware of this, and it's good to know.

 

[518er2005]

Hi, I did as you said, but still getting the same error. It creates the init.nc file perfectly, but when I am trying to create the lbc conditions am getting this error. By the way, is it necessary that we need to run in four mpi tasks?

Cheers,
Debashis

 

[mgduda]

When you say "the same error", do you mean the error "execvp error on file init_atmosphere_model (No such file or directory)"? If so, you could try adding the relative or absolute path of the init_atmosphere_model executable in your mpiexec command; for example, use "./", as in

mpiexec -n 4 ./init_atmosphere_model

In any case, it's not necessary to run with four MPI tasks, but you'll likely want to use more than one MPI task when running the model. Since the init_atmosphere_model and atmosphere_model programs use the same parallel infrastructure, then, sorting out the issue with the init_atmosphere_model program would likely be time well spent.

 

[518er2005]

Thanks, it worked and it's running perfectly. And ya there is a yum package for metis which works as wonder.

Cheers,
Debashis

 

[mgduda]

Thanks for following up! It's good to hear that adding "./" worked.