You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Currently the hotstart generator has as far I can tell, only one "date" field. It generates a hotstart with a nominal time=0. This makes it hard to use particularly for runs that pause at an intermediate date and restart from there with a new grid (island flooding is an example). We have also run into this issue on some calibration runs for the South Delta where the grids were evolving and we wanted to a test a particular mid-run date. The consequences of not realizing that this deficit exists and trying to restart is that the run will restart from time=0 (which is the date in the hotstart) and all the staout_* and flux.out files will be overwritten. So if you have a misunderstanding, it is possible you will lose data and not be able to recover without a complete do-over.
The feature request here is to add a required additional field that has a trivial default:
date: 2021-04-20
run_start: default # clarfies this is the same as "date" which is appropriate for ihot=1
run_start: 2021-01-01T00:00 # if the simulation has a nominal start of 2021-01-01 and is being restarted on 2021-04-20.
There is currently a hack script hotstart_date.py to allow assignment of a date. It may be a useful reference. The fields "time", "nsteps_from_cold","iths" and "ifile" need to be filled. I'm not sure about what ifile does, so please look into that. It would also be nice if "run_start" could be added as an (unrequired) attribute. In the script I used "time_origin_of_simulation".
It would also be nice if there were a flag, default True, that would use the hotstart file name in the way the script does ... so that hotstart.nc is changed to:
hotstart.20210420.1234567.nc
I don't see any reason that you couldn't arrange it so the dt changes, but we can make that a separate issue.
The text was updated successfully, but these errors were encountered:
Currently the hotstart generator has as far I can tell, only one "date" field. It generates a hotstart with a nominal time=0. This makes it hard to use particularly for runs that pause at an intermediate date and restart from there with a new grid (island flooding is an example). We have also run into this issue on some calibration runs for the South Delta where the grids were evolving and we wanted to a test a particular mid-run date. The consequences of not realizing that this deficit exists and trying to restart is that the run will restart from time=0 (which is the date in the hotstart) and all the staout_* and flux.out files will be overwritten. So if you have a misunderstanding, it is possible you will lose data and not be able to recover without a complete do-over.
The feature request here is to add a required additional field that has a trivial default:
date: 2021-04-20
run_start: default # clarfies this is the same as "date" which is appropriate for ihot=1
run_start: 2021-01-01T00:00 # if the simulation has a nominal start of 2021-01-01 and is being restarted on 2021-04-20.
There is currently a hack script hotstart_date.py to allow assignment of a date. It may be a useful reference. The fields "time", "nsteps_from_cold","iths" and "ifile" need to be filled. I'm not sure about what ifile does, so please look into that. It would also be nice if "run_start" could be added as an (unrequired) attribute. In the script I used "time_origin_of_simulation".
It would also be nice if there were a flag, default True, that would use the hotstart file name in the way the script does ... so that hotstart.nc is changed to:
hotstart.20210420.1234567.nc
I don't see any reason that you couldn't arrange it so the dt changes, but we can make that a separate issue.
The text was updated successfully, but these errors were encountered: