Problem with compiling the MPAS init_atmosphere_model

[Miftha Kurnia Putri]

Dear All,

I am attempting to run the init_atmosphere_model in MPAS with the 10km_uniform configuration. However, during execution, the process unexpectedly stops and is killed. Upon checking the log.init_atmosphere.0000.out file, I found no errors or any output related to static.nc. What steps should I take to resolve this issue?

I have also attached the streams.init_atmosphere, namelist.init_atmosphere, and log.init_atmosphere.0000.out files. Please assist me in solving this problem.

Thank you very much.

Sincerely,

 

[Ming Chen]

You didn't include namelist.init_atmosphere in your zip file. Please post your file for me to take a look.

Also, it seems that you run this case with only a single processor. if so, can you retry with more processors? I am concerned that memory could be an issue.

 

[Miftha Kurnia Putri]

This is the first time I am running the MPAS model, and I am still learning how it works. Could you please tell me how to check the processors I have and how to run it with more processors?

i included the namelist.init_atmosphere file.

 

[gdicker]

how to check the processors I have

It is difficult to answer this question without knowing the system you are trying to run on. Assuming you are on a Linux system (e.g. Ubuntu), you should try lscpu on the command-line and look through the output for a line like "Core(s) per socket: <SOME_NUMBER>". If you are on another system (or lscpu doesn't work for you), I recommend you use Google to help find the answer.

how to run it with more processors?

Since MPAS uses MPI (message passing interface), you can run with multiple processors by using a command like:

mpiexec -n 4 ./init_atmosphere_model

As Ming mentioned, the issue may be the amount of memory available on your system. I would recommend using Google to find the commands to figure out how much memory/RAM your system has.

 

[Ming Chen]

Hi,
(1) In your namelist, you set config_block_decomp_file_prefix = 'x1.40962.graph.info.part.'. However, your streams.init_atmosphere file indicates that you are processing the data on the mesh of "x1.5898242.grid.nc". therefore, you need to create graph file for parallel run over the x1.5898242.grid mesh.
(2) To create files for parallel run, you need to install METIS. Detailed descriptions can be found in https://www2.mmm.ucar.edu/projects/mpas/mpas_atmosphere_users_guide_8.1.0.pdf (see Section 4.1)
(3) Please read the MPAS Users Guide to gain a complete picture how MPAS works.
(4) There is an online MPAS practice tutorial, which will guide you through the entire process step-by-step. Please see the link here.
Hope these are helpful for you. Let us know if you have any issues during the process.

 

[Miftha Kurnia Putri]

Dear all,

lscpu

I tried the command line and it said that i have 6 cores, Thanks!.

To create files for parallel run, you need to install METIS

I do the instruction to install METIS and do the command "gpmetis -minconn -contig -niter=200 x1.5898242.graph.info 512". I also changed the set config_block_decomp_file_prefix = 'x1.5898242.graph.info.part.512'.

When I running it with mpiexec -n 4 ./init_atmosphere_model command, it appear the same error. Please help me why it still fails..

 

[Ming Chen]

There are a few issues here:
(1) Please modify your namelist.init_atmosphere with the following option:

&decomposition
    config_block_decomp_file_prefix = 'x1.5898242.graph.info.part.'
/

(2) For this large case with 5898242 meshes, I would suggest that you create graph file for using more processors, e.g.,
gpmetis -minconn -contig -niter=200 x1.5898242.graph.info 2560
(3) when you run the case, the number of processors you use should be consistent with the number shown in your graph file. For example, you should run with 512 processors if you have x1.5898242.graph.info.part.512 in your working directory.
(4) Your streams file look fine to me.

Please try and let us know if you still have issues.

 

[Ming Chen]

I would suggest that you start from a small case, for example the case of 240-km mesh given in the MPAS practice website. A small case can run quickly, yet it covers all the necessary information and steps you need to know about MPAS run. I suppose this is a more efficient way to learn MPAS.

 

[Miftha Kurnia Putri]

Hello, thank you for your reply.

I am currently attempting the MPAS practice tutorial using a 240 km mesh with data sourced from NCAR RDA Dataset ds084.1. I have processed this data in WPS to produce met_em.nc, which I intend to use in namelist.init_atmosphere. However, this has resulted error on reading namelist record nhyd_model. Does the MPAS model only support data from ERA5 and CFSR?

 

[Ming Chen]

The intermediate file produced by ungrib program is used as input to produce MPAS initial condition. You shouldn't use met_em file.

Again, please follow the instructions and steps in MPAS practice website, which is helpful for you to learn MPAS.