Hello, everyone. I am new to use CMAQ model. I am trying to do a simulation based on data from “2020 Modeling Platform with model ready and pre-merged emissions, icbc and mcip for running CMAQ”. I am trying to do a two day simulation based on the script(I have modified them) offered in the amazon data ware house. I cannot run it and i cannot find efficient information from the log file. I hope someone can help me. Thank you in advance! The attachments are the data structure, run scripts and the log files I have got. If you need more information, please let me know CTM_LOG_001.2020ha2_MP_cb6r5_ae7.v54.BE.12US1_35_20200801.txt (18.1 KB)
Hi,
I did download the LOG file and checked through. However, I didn’t find anything wrong for the LOG file that you uploaded. Please try to upload all LOG files that you have then we can figure it out as sometimes error messages pop up in other LOG files.
It is not clear how you are submitting this job, but it seems you are using slurm. You are telling the model to run on 2 nodes for a total of 88 processors. Normally we would set the number of processors to be equal to NPROCS, the product of NPCOL and NPROW. What is the content of this log file:
/proj/ie/proj/CMAS/CMAQ/2020_12US1/CMAQ_v54+/CCTM/scripts/cmaq_%j.txt
where the % is replaced by the slurm job ID.
Have you run one of the benchmark tutorial cases yet? If not, I strongly suggest doing so, to ensure the system is set up correctly, particularly regarding running in parallel.
Hi, cgnolte. I will check the log file. I have run the 2018 U.S. benchmark two day test case successfully. I submitted the task in the WSL in my own PC which has only 8 cores in the CPU. The lines you mention I found it commented in the sample so I ignored them. I will take a look of them. Thanks for your suggestion.
To try to make sure the problem is parallel, I have tried to use serial to run the case. The simulation still stock in this step. I think the problem may not be parallel computation.
Do you mean you are trying to run CMAQ on a Windows PC? I have not done this, and would not recommend it.
You have to make sure that you have an environment that supports MPI. You should be able to find a simple MPI test program.
I would not try to run a 299x459x35 domain on a single processor. That will take an unreasonably long time, even if you have enough memory for it to run.
Another problem that can occur on Windows machines is the machine goes into sleep mode when the user stops interacting with it. You have to make sure that sort of power management is turned off.
Yes, I am using a virtual machine in Windows to try to do that. Ok, I can understand that the large domain may be a problem for this. For the MPI, I just went through the benchmark tutorial and get the result. Thanks for your suggestion! I may need to get a native Linux machine for the simulation.