NTASK configuration for CESM2.1.3 pacemaker experiment

[chunlienchiang0113]

Hello,

I tried to create a pacemaker expeirment in CESM2.1.3 based on the instruction here (CESM2 Pacific Pacemaker Ensemble Instructions | Community Earth System Model), and the simulation will be failed under some specific CPU configurations.

I run the experiment on my computer, and the following configuration is attempted:

1. 320 cores for all components: successful
2. 480 cores for atm, cpl, ocn, glc and 320 cores for other components: successful
3. 480 cores for atm, cpl, ocn, glc, wav and 320 cores for other components: failed (idling for more than several hours without updating log files)
4. 320 cores for wav and 560 cores for other components: failed (raise error during mapper_Sw2o initialization)
5. Default BHIST with any normal configuration (320 or 1024 for all components): successful

I would like to ask if anyone has ever encountered the problem that the pacemaker experiment in CESM2.1.3 has specific CPU configuration requirements?

The attached zip file is the log files and env_mach_pes.xml file for the failed run (4) mentioned above.

Thanks!

 

[jedwards]

I'm surprised that the wav model works on as many as 320 cores - I would limit it to use only one node.
You can do that with ./xmlchange NTASKS_WAV=-1