Update to spack-stack 1.6.0#2239
Conversation
|
@malloryprow FYI I am starting to test this in cycled mode. Do you have an EMC_verif-global branch you would like me to include? |
Awesome! Yes, it is https://github.com/NOAA-EMC/EMC_verif-global/tree/gw_updates, commit NOAA-EMC/EMC_verif-global@1e7baf7 |
|
@CoryMartin-NOAA @RussTreadon-NOAA Would you like to add an updated GDASApp hash that includes spack-stack 1.6.0 support to this PR? I gave it a quick try myself, but there are a couple of libraries that are not available in 1.6.0, so I'm not sure if these need to be requested or if there are appropriate substitutes: My attempt is here: |
|
It would be nice to get GDASApp to spack-stack/1.6.0 but I'm not sure if JEDI repos are ready for this. What do you think @CoryMartin-NOAA . @DavidHuber-NOAA , I will look at your |
I'm not sure. I have no reason to suspect they aren't ready but I also haven't seen JCSDA migrate to using 1.6 for Skylab yet. |
|
While I was able to build GDASApp Rebuild |
|
The failed followed by Not sure where to go with this but I'll keep poking around. |
|
Looks like the python version used by the ctests is not the correct one. @guillaumevernieres had a similar issue with his AI tinkering on Hera. We may need to do the following in the cmake command: ``-DPYTHON_EXECUTABLE=`which python``` |
|
I made the following change to Executing However, later on in the build I see Looks like I need to add |
|
Hmm, I'm pretty sure it is because the |
|
Unloading Rerun |
|
Forgot to add the following to my Hera environment With the above set, only one test fails I am still concerned about the other modules unloaded when |
|
@malloryprow The |
How exciting to see stats being produced in the global-workflow again! The files look fine! Could I see the log files? I want to check on something. |
|
@malloryprow Fantastic! Sure, the log files are here: |
Me too, @RussTreadon-NOAA . I think this is driving me to want to replace the python env with a new one that uses spack-stack python instead of hpc-stack miniconda, but that won't be trivial. The trivial hack fix is to put in the modulefile a prepend to the PYTHONPATH var that looks in our numpy install first. |
Thanks! I was curious as where the connector script was pointing model_dir_list for EMC_verif-global to. The default value in run_verif_global_in_global_workflow.sh#L36 kicked in since model_dir is not defined in config.metp. It would be good to get model_dir defined in config.metp. Would that be okay for this PR? It be adding one line in config.metp on line 24, like in https://github.com/malloryprow/global-workflow/blob/4acc025cdcacdba4ac85c647420a5577854deac3/parm/config/gfs/config.metp. |
|
@malloryprow Sure, I can put that in. I'll let it run through another cycle so you can check the logs and |
|
@CoryMartin-NOAA and @DavidHuber-NOAA , I added Not sure if this is the correct solution. |
|
@malloryprow Alright another GFS cycle completed. Could you check the |
|
@RussTreadon-NOAA if it builds, and if the 'workflow tests' pass, i.e. the ones that run the JEDI executables, then I think it might be fine. |
I see model_dir defined through config.metp in all 3 logs (gfsmetpg2g1.log, gfsmetpg2o1.log, gfsmetppcp1.log), and the stat files look good. Since everything looks good, I would like to merge the branch these changes are in EMC_verif-global:gw_updates into EMC_verif-global:develop. I think that may need to update the branch or commit that the global-workflow is checking out? Is that right @DavidHuber-NOAA? |
|
|
99.9% sure this commit would solve this issue: 8cbc974 |
|
Yes, I agree with Cory. This failure was caused by recent merged commit hash by changing from |
|
Per suggestion from @aerorahul I'm going to proceed without running the snow DA tests. We can reenable them when the issues have been sorted out in a future PR. |
|
|
This is caused by the following commit hash. Cory's new PR #2330 will solve this issue. For now, you can use the early commit hashes as below for your test. |
|
Automated global-workflow Testing Results: |
|
Experiment C48_ATM_79d305e8 SUCCESS on Hera at 02/22/24 11:04:08 PM |
|
Experiment C48_S2SW_79d305e8 SUCCESS on Hera at 02/22/24 11:48:10 PM |
|
Experiment C96C48_hybatmDA_79d305e8 SUCCESS on Hera at 02/23/24 12:16:19 AM |
|
Experiment C96_atm3DVar_79d305e8 SUCCESS on Hera at 02/23/24 12:16:26 AM |
|
All CI Test Cases Passed on Hera: |
Remove `FIX*` variables for fix subfolders and replace them with the remaining `FIXgfs` variable and the subfolder name (e.g. `${FIXam}` -> `${FIXgfs}/am`).
The UFS_UTILS and GDASApp repos were similarly updated. This PR includes a new UFS_UTILS hash.
The updated GDASAPP hash was already committed within the spack-stack/1.6.0 PR #2239.
Resolves #2184
Description
This updates the module files for the global workflow to use spack-stack 1.6.0 and upgrades the submodules
(except GDAS)to the same versions. This also addresses a couple minor issues discovered during testing: a bug in theawips_g2job wherefcsthrswas undeclared,build_upp.shcan now accept the-j #flag to specify the number of build jobs, an optimization was made tobuild_all.shto build the UFS with more build jobs which speeds up the overall build process, and inconsistencies betweenbuildandrunversion files were corrected.This update allows two major upgrades to take place simultaneously:
With spack-stack v1.6.0, the name of the gsi-addon environment is different across platforms, so the version files have been restructured to construct the path to the
Coremodule files.The libraries that were upgraded in this upgrade that affect the global workflow are
crtm 2.4.0 -> 2.4.0.1
netcdf-fortran 4.6.0 -> 4.6.1
cdo 2.0.5 -> 2.2.0 (except on Orion)
prod_util 1.2.2 -> 2.1.1
py-yaml 5.4.1 -> 6.0.0
xarray 2022.3.0 -> 2023.7.0
The new libraries used in the global workflow to support EMC_verif-global are
pandas v1.5.3
python_dateutil v2.8.2
met v9.1.3
metplus v3.1.1
Resolves #2195 #1734 #1575
Depends on
Upgrade spack-stack from version 1.5.1 to 1.6.0 ufs-community/ufs-weather-model#2093The UFS upgrade will be tackled at a later time.Type of change
Change characteristics
How has this been tested?
Checklist