ESCOMP / CISM-wrapper

Community Ice Sheet Model wrapper for CESM
http://www.cesm.ucar.edu/models/cesm2.0/land-ice/
Other
3 stars 15 forks source link

tavg_helper file should possibly be added to config_archive.xml #68

Open billsacks opened 2 years ago

billsacks commented 2 years ago

As I'm moving CISM's config_archive.xml entries into this repository (under cime_config/config_archive.xml) and reworking it to support multiple ice sheets, it occurred to me that the tavg_helper file should possibly be listed here. This could be important if the tavg_helper file should be saved to the archive and if it's important for this file to be retrieved from the archive and restaged in the run directory in order for restarts / branch runs to be bit-for-bit in the diagnostic fields that this file is tracking when a run stops part-way through an averaging period.

Since I don't fully understand the tavg_helper file, I'm not positive that this is necessary (though I suspect that it is under some conditions). If this is added to config_archive.xml, however, we may run into the same issue we ran into for the initial_hist file: the hist_file_extensions in config_archive are also used to tell the test system which files to compare in system tests, which can possibly cause problems in restart test comparisons. If this is an issue, we may be able to deal with it by either:

  1. Listing this as a rest_file_extension instead of hist_file_extension. This would need some thought / investigation about which is appropriate here... it may actually be that rest_file_extension is most appropriate. (I think the differences are in whether files are moved or copied to the archive, retrieved from the archive, and whether only the last time's file is dealt with or all times; there may be other differences, too.)
  2. Putting some special purpose code in cime to ignore these helper files in tests (that's how we ignore initial_hist files, though that isn't ideal)
  3. (I thought another solution may be to use the rest_history_varname option in config_archive, but I think that's probably not the right thing to do here.)

@Katetc @whlipscomb @gunterl - fyi.