- Log in to post comments
I'm trying to run HWRF on NOAAs Orion supercomputing cluster and have been running into a few issues. I've been able to troubleshoot most things along the way, but was wondering if there are any SLURM batch scripts available to work from like there are with Cheyenne and MOAB/torque? Or at least someone who has any guidance on what I may be doing wrong?
In particular, when running the HWRF forecast, I am getting slow performance (~1 hour of wall time for every 3 hours of forecast), leading to a timeout when using the suggested resources in the wrapper script forecast_wrapper. I've tested what happens when I reduce the forecast to 15 hours, and while I get through the forecast, something seems to hang at the end and eventually time out. The last few lines of my output log are below:
starting wrf task 5 of 660
starting wrf task 9 of 660
starting wrf task 7 of 660
starting wrf task 3 of 660
coor_mask_a nide= 1165 njde= 2335
coor_mask_a read nhlat 2720275 words
nhlat= -19.4872639487162 -19.4487278598845 -19.4234890004944
coor_mask_a read nhlon 2720275 words
coor_mask_a read nvlat 2720275 words
coor_mask_a read nvlon 2720275 words
coor_mask_a read hres_sm 2720275 words
srun: Job step aborted: Waiting up to 32 seconds for job step to finish.
I've looked through a number of the other logs (can provide any as needed) and can't seem to find anything that would indicate what is going awry, so any input on which SLURM directives to use and/or model configurations would be greatly appreciated.
