Conversation
compiles and starts, but misses (as expected) information in existing JSBACH restart files. Cold start not yet implemented
links all currently known files correctly running not yet tested as info from Shradda is still missing
draft version of ECHAM76-LMU that start up correctly but still fails due to land sea mask mismatch
fixed restart setup in a hacky way (see foci.yaml) added a new runscript for FOCIOIFS
…geomar_dev exludes the conflicting oasis.py
This should be the only difference between geomar_dev and release that matters
allow different streams for restart and output for ECHAM and JSBACH
with FOCI-MOPS with the 12 tile edition
merge geomar dev into geomar branch
separate echam6 namlist templates for foci and foci_lmu
before merging this branch into geomar_dev differences where too large for an efficient direct merge
example runscript (foci-default) on how to use input from lustre-rzg on Emmy p3
This reverts commit a92860b.
made the new namelists default for an oasismct4_mops based setup
adds dask stuff, moves post run commands to computer section
mandresm
left a comment
mandresm
left a comment
There was a problem hiding this comment.
First round of minor requests and questions :)
| streams: | ||
| - echam | ||
| - accw | ||
| - co2 | ||
| - g3bid | ||
| - g3bim | ||
| - g3bday | ||
| - g3b1hi | ||
| - glday | ||
| - aclcim | ||
| - sp6h | ||
| - glim | ||
| - spim | ||
| - ism | ||
|
|
||
| restartstreams: | ||
| - echam | ||
| - accw | ||
| - co2 | ||
| - g3bid | ||
| - g3bim | ||
| - g3bday | ||
| - g3b1hi | ||
| - glday | ||
| - aclcim | ||
| - sp6h | ||
| - glim | ||
| - spim | ||
| - ism | ||
|
|
There was a problem hiding this comment.
Hi @christian-stepanek, @chrisdane, @PengyangSong and @ackerlar,
In this PR we are merging efforts from GEOMAR team into release. This is important because it will make the whole thing more maintainable for them as well as for us (and because they have put quite some work to add ICON-FESOM ;) ). Before the PR is merged we will check that no model behavior is affected (as we usually do).
However, here is a change in how things are done in ECHAM until now, and I want to get your approval and keep you in the loop. In the past the list of streams, referenced above, was used both to compute the name of the output files and the restart files. However, it might be the case that some times some streams have restarts but do not have outputs, or the other way around. What @seb-wahl did here (and in the rest of the file) was to duplicate the the list of streams into a new list restartstreams and use the second one for all the operations that affect restart files. That does NOT lead to any change in behavior as the 2 lists are identical by default. However, this allows you to add new streams from your runscript to either streams or restartstreams so you can now you could have control of output and restarts independently if you need to.
If you are okay with this change please give a thumbs up to this message, if you have concerns please reply :)
There was a problem hiding this comment.
@chrisdane, since you work with streams quite a bit and modify them, can you answer to this?
mandresm
left a comment
mandresm
left a comment
There was a problem hiding this comment.
Reviews, comments and todos after the merge party
| glob_search_file = ( | ||
| f"{restart_file_path}*" | ||
| f"{config['ini_restart_dir']}{restart_file}*" | ||
| f"{config['ini_restart_date'].year}" |
There was a problem hiding this comment.
@mandresm test these changes with restarting from the pool
There was a problem hiding this comment.
Ok, you know FESOM better than I do so just use copy instead. Optimization can be done later if necessary.
There was a problem hiding this comment.
@seb-wahl, this reply refers to this comment of mine #1443 (comment) right?
There was a problem hiding this comment.
I run some bit-identical manual tests for restarts for AWICM3. For some of them, restarts showed some non-bit-identical files in rstos and rstas (coupling files for awicm3) between the release branch and the geomar_prep_release, butcdo diffv` showed that the data is identical (with CDO_TOLERANCE=0). For those specific problematic tests, the rstos and rstas files were created in that run, and it is very likely that they are created differently depending on the order of the fluxes that are output, but the data is identical. In fact, for those tests fesom restart files are identical between release and geomar batch for all runs.
For all the other scenarios (branching-off from a parent experiment and starting a run with oasis3mct.lresume: True) rstos and rstas where bit-identical and the behaviour was as expected.
All runs showed differences in LAW and srf oifs restart files, but I think that's again expected due to mpi-task ordering.
There was a problem hiding this comment.
Is there t2m or sst output for that day of simulation? If so, could you plot and show it?
for oifs it should work with ncview. for fesom with ushow
There was a problem hiding this comment.
Here the plots of the vars you asked (I think). One is for release (rf0) and another for the geomar_prep_release branch (gf0).
/work/ab0246/a270152/gf0/outdata/oifs/atm_remapped_1d_2t_1d_2000-2001.nc_20000201-20000229
/work/ab0246/a270152/rf0/outdata/oifs/atm_remapped_1d_2t_1d_2000-2001.nc_20000201-20000229
/work/ab0246/a270152/gf0/outdata/fesom/sst.fesom.200002.01.nc
/work/ab0246/a270152/rf0/outdata/fesom/sst.fesom.200002.01.nc
diff /work/ab0246/a270152/gf0/outdata/fesom/sst.fesom.200002.01.nc /work/ab0246/a270152/rf0/outdata/fesom/sst.fesom.200002.01.nc yields identical results
Cdo diff of the grib files differ :/ any clues @JanStreffing?
$ cdo diffv /work/ab0246/a270152/gf0/outdata/oifs/atm_remapped_1d_2t_1d_2000-2001.nc_20000201-20000229 /work/ab0246/a270152
/rf0/outdata/oifs/atm_remapped_1d_2t_1d_2000-2001.nc_20000201-20000229
cdi warning (find_time_vars): Found more than one time variable, skipped variable time_centered!
cdi warning (cdfInqContents): Coordinates variable time_centered can't be assigned!
cdi warning (find_time_vars): Found more than one time variable, skipped variable time_centered!
cdi warning (cdfInqContents): Coordinates variable time_centered can't be assigned!
Date Time Level Gridsize Miss Diff : S Z Max_Absdiff Max_Reldiff : Parameter name
1 : 2000-02-01 12:00:00 0 76800 0 37239 : F F 1.5804 0.0064408 : 2t
2 : 2000-02-02 12:00:00 0 76800 0 37251 : F F 1.5486 0.0063757 : 2t
3 : 2000-02-03 12:00:00 0 76800 0 37253 : F F 1.6022 0.0065470 : 2t
4 : 2000-02-04 12:00:00 0 76800 0 37248 : F F 1.4216 0.0057091 : 2t
5 : 2000-02-05 12:00:00 0 76800 0 37243 : F F 1.0744 0.0038981 : 2t
6 : 2000-02-06 12:00:00 0 76800 0 37243 : F F 1.2413 0.0044807 : 2t
7 : 2000-02-07 12:00:00 0 76800 0 37236 : F F 1.0529 0.0038130 : 2t
8 : 2000-02-08 12:00:00 0 76800 0 37248 : F F 1.0907 0.0043480 : 2t
3 participants
This PR finally brings GEOMAR's contributions to ESM-Tools back into release. It will allow the release branch of ESM-Tools to correctly handle