Aurora_LagosD
New member
Hi everyone,
My collagues and I have been trying to run WRF 4.5.1 in a HPC. We've noticed that when we use the MYNN2.5 scheme for pbl an error occurs. The simulations stops within its first 10 seconds before any timestep calculation even happens. Has anyone reported this before?
1. We are using the attached namelist.input.Reference when we get this error.
2. We've also ran the attached namelist.input.Test successfully.
3. Hence, we've modified the physics flags one by one of the Test case to match the Reference case until we got an error. The error occurs ONLY when we change the bl_pbl_ physics to the MYNN 2.5 scheme.
4. We've tried running bl_pbl_physics = 5 with sf_sfclay_physics=1, 2 and 5. All of these combinations produce the same error.
5. We've ran the job in different partitions, with different CPU memory capacity, sequencially, with OpenMP, MPI and MPI-OpenMP and all of them produce the same error.
6. We've contacted other colleagues for them to run their local installation of WRF 4.5.2 with our namelist and they've encountered the same situation. That is bl_pbl_physics=1 runs successfully but when it is changed to 5 it crashes.
7. We've also tried with different meteorological boundaries fields and geostationary data, but the same outcome always happens, when we use bl_bpl_option=5 it crashes and when we use 1 it's successful .
We must use the MYNN2.5 scheme because we have to quantify the benefits of a change from a current set-up in WRF 3.6 to a newer set up in WRF 4.5.2.
We need support at this point to run WRF 4.5.1, we've exhausted all possible solutions that we know to look for to run the namelist.input.Reference. Any suggestions are appreciated.
My collagues and I have been trying to run WRF 4.5.1 in a HPC. We've noticed that when we use the MYNN2.5 scheme for pbl an error occurs. The simulations stops within its first 10 seconds before any timestep calculation even happens. Has anyone reported this before?
1. We are using the attached namelist.input.Reference when we get this error.
2. We've also ran the attached namelist.input.Test successfully.
3. Hence, we've modified the physics flags one by one of the Test case to match the Reference case until we got an error. The error occurs ONLY when we change the bl_pbl_ physics to the MYNN 2.5 scheme.
4. We've tried running bl_pbl_physics = 5 with sf_sfclay_physics=1, 2 and 5. All of these combinations produce the same error.
5. We've ran the job in different partitions, with different CPU memory capacity, sequencially, with OpenMP, MPI and MPI-OpenMP and all of them produce the same error.
6. We've contacted other colleagues for them to run their local installation of WRF 4.5.2 with our namelist and they've encountered the same situation. That is bl_pbl_physics=1 runs successfully but when it is changed to 5 it crashes.
7. We've also tried with different meteorological boundaries fields and geostationary data, but the same outcome always happens, when we use bl_bpl_option=5 it crashes and when we use 1 it's successful .
We must use the MYNN2.5 scheme because we have to quantify the benefits of a change from a current set-up in WRF 3.6 to a newer set up in WRF 4.5.2.
We need support at this point to run WRF 4.5.1, we've exhausted all possible solutions that we know to look for to run the namelist.input.Reference. Any suggestions are appreciated.