-
Notifications
You must be signed in to change notification settings - Fork 258
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Refactor fv3atm history & restart to reduce redundant code. Add rrfs-sd and clm lake to quilt restart. #1769
Refactor fv3atm history & restart to reduce redundant code. Add rrfs-sd and clm lake to quilt restart. #1769
Conversation
…estart into fv3atm_restart_io.F90
…ufs-weather-model into refactor-restart
…ufs-weather-model into refactor-restart
…ufs-weather-model into refactor-restart
@BrianCurtis-NOAA I think I had a similar issue on Jet, and I had to recreate the baselines again for those failing cases, then re-copy them over. |
Seemed to be OK this time. I know Acorn is a test machine, but still concerned about seeing this issue again. Can address later if we do see it again. |
Looks like all the testing is complete, so we can now begin the merge process. |
Ok, the FV3 subcomponent PR was merged in, please go ahead and revert gitmodules and update hash. |
I have reverted the gitmodules and updated the hash. You may proceed with the merge. |
I just noticed the conus13km cases are not actually doing a quilt restart (even the ones with qr) because of a missing line in a model_configure file. I am testing now if that functionality works. The regional domain quilt restart was not the intent of this PR, so I think this PR can be merged anyway. However, if you want to go ahead with the merge, I suggest you disable those tests to save CPU time. That's these two:
I will fix them in my next PR (32-bit physics in RRFS) if they don't already work. |
You can also leave rt.conf how it is, and you'll just run two tests that are duplicates of older ones. Please let me know what you want to do. |
@SamuelTrahanNOAA I'll go ahead and disable those two in the meantime, thanks |
e70b591
I have disabled those four tests in rt.conf. |
I can happily report those two tests do pass when I add the missing line to model_configure_rrfs_conus13km.IN. I have made this change to my other PR already. diff --git a/tests/parm/model_configure_rrfs_conus13km.IN b/tests/parm/model_configure_rrfs_conus13km.IN
index 5be098ed..6feac2c0 100644
--- a/tests/parm/model_configure_rrfs_conus13km.IN
+++ b/tests/parm/model_configure_rrfs_conus13km.IN
@@ -12,6 +12,7 @@ restart_interval: @[RESTART_INTERVAL]
output_1st_tstep_rst: .false.
quilting: @[QUILTING]
+quilting_restart: @[QUILTING_RESTART]
write_groups: @[WRITE_GROUP]
write_tasks_per_group: @[WRTTASK_PER_GROUP]
output_history: @[OUTPUT_HISTORY] |
@SamuelTrahanNOAA that's great to hear, thanks for verifying that those tests are passing and good to go, I'll go ahead and finish up on merging this PR. |
Description
This is a major refactoring of the sfc restart code. It eliminates code duplication between read vs. write and quilt vs. non-quilt. Most notably, it merges the sfc read & write grid copy loops for quilt and non-quilt into a single function. With these changes, the code will be easier to maintain, and read vs. write bugs will be easier to find. See NOAA-EMC/fv3atm#660 for full details on that.
You'll see several new quilt restart regression tests, including 32-bit physics, clm lake, and rrfs-sd. Those new tests are also run for the gnu compiler, so we'll know the quilt restart will continue to work for the non-Intel world. I've also replaced one RRFS restart test. It was run without smoke, but should have been run with smoke.
There's a fix for another bug, #1781, wherein nccmp generates many gigabytes of output when fields differ. The fix is to add
-q -S
. This disables printing of data and metadata differences, but enables a table of statistics comparing the data.Also, a fix for but #1780: when rt.sh aborts due to another rt.sh running, it deletes the
lock
directory. That directory should only be deleted by the rt.sh that created it. This happens due to an error in the script's error handler.Top of commit queue on: TBD
Input data additions/changes
Anticipated changes to regression tests:
I had to generate baselines for these tests on Hera:
Expand for full list.
Subcomponents involved:
Combined with PR's (If Applicable):
Commit Queue Checklist:
Linked PR's and Issues:
Component PRs:
Related issues:
-f
option tonccmp
#1781Testing Day Checklist:
Testing Log (for CM's):