-
Notifications
You must be signed in to change notification settings - Fork 2
Common problems
mikebedington edited this page Jun 14, 2022
·
3 revisions
If things go wrong some common problems are:
- It is nearly always the GFS download (for the WRF model) which goes wrong; either because its hanging on a particular file or its had too many rejections so has skipped downloading for several days in row which leaves a gap it can't fill. Both these need the download suite restarting for a date starting a few days prior. The easiest way to see whether this is the issue is to look at the rolling data on ceto '/gpfs1/users/modellers/modop/Data/GFS_rolling'.
- A fall over in adjust_soil_levels or met_grid is often actually that the GFS download didn't run properly; you can check the dates its expecting in metgrid.log and the dates avaiable in /gpfs1/users/modellers/modop/Data/GFS_rolling
- Recently there was a numerical error in the actual running of WRF; I solved this by reducing the timestep of the model. This is in apps/write_run_namelist/file/namelist.input the time_step variable which I changed from 100s to 60s.
- Check the amount of disk space on ceto and sthenno as often it has just maxed this out
- Occasionally it crashes with NaNs, I think this is probably boundary condition related and usually is solved by restarting the run using the 'COLD_START=True' option (technically not a true cold start since it generates the initial conditions from CMEMS)
- It had a random outage recently which I think was related to the discs on ceto getting a bit old...