Masiello_Cameron
New member
Hey,
I am currently trying to run WRF on NCAR derecho in em_real, with a storm following inner nest. I have a quiescent initial condition environment, to which I then add a bogus vortex. So I have two wrf_input d01 files, one with a bogus vortex inserted into the wrf_input file and one without. I am using the pre-compiled version of WRF version 4.5 that is on derecho, and I re-compiled it using the storm following configuration of em_real.
If I run em_real with the wrf_input without a bogus vortex added, it runs fine throughout the 192-hour run time, with regular convective evolutions as expected.
However, when I run with the bogus vortex, the simulations starts fine before crashing randomly 2.5 hours into the simulated run. There are no apparent errors as to why my run crashes, but these exit codes are outputted: "rank 56 exited with code 174, rank 60 exited with code 1." There are no CFL violations when I grep through the out files, and the wrfout files look mostly fine with no apparent noise propagation occurring along the inner nest or parent domain boundaries.
I tried running with DFI turned on, thinking maybe the bogus vortex was creating some mayhem in the model, but that sped up the time of crashing to an hour and 50 minutes into the simulated run.
To add further confusion to the problem, I ran these same simulations (the same met_em files, input files, wrf_bdy files, version of WRF, and nest following) on a local Linux cluster, and they ran fine without crashing regardless of the wrf_input version utilized (bogus and non_bogus).
I attached the rsl.out and rsl.error files from the derecho runs with the namelist I am utilizing. I would greatly appreciate any advice you could give me.
I am currently trying to run WRF on NCAR derecho in em_real, with a storm following inner nest. I have a quiescent initial condition environment, to which I then add a bogus vortex. So I have two wrf_input d01 files, one with a bogus vortex inserted into the wrf_input file and one without. I am using the pre-compiled version of WRF version 4.5 that is on derecho, and I re-compiled it using the storm following configuration of em_real.
If I run em_real with the wrf_input without a bogus vortex added, it runs fine throughout the 192-hour run time, with regular convective evolutions as expected.
However, when I run with the bogus vortex, the simulations starts fine before crashing randomly 2.5 hours into the simulated run. There are no apparent errors as to why my run crashes, but these exit codes are outputted: "rank 56 exited with code 174, rank 60 exited with code 1." There are no CFL violations when I grep through the out files, and the wrfout files look mostly fine with no apparent noise propagation occurring along the inner nest or parent domain boundaries.
I tried running with DFI turned on, thinking maybe the bogus vortex was creating some mayhem in the model, but that sped up the time of crashing to an hour and 50 minutes into the simulated run.
To add further confusion to the problem, I ran these same simulations (the same met_em files, input files, wrf_bdy files, version of WRF, and nest following) on a local Linux cluster, and they ran fine without crashing regardless of the wrf_input version utilized (bogus and non_bogus).
I attached the rsl.out and rsl.error files from the derecho runs with the namelist I am utilizing. I would greatly appreciate any advice you could give me.