Build ufs_weather_model on Gaea-C5 + update cubed sphere gitmodules for perturbation/increments for cold starts #2271

[DavidBurrows-NCO]

Commit Queue Requirements:

• Fill out all sections of this template.
• All sub component pull requests have been reviewed by their code managers.
• Run the full Intel+GNU RT suite (compared to current baselines) on either Hera/Derecho/Hercules
• Commit 'test_changes.list' from previous step

---

Description:

• Updates modulefiles/ufs_gaea.intel.lua to compile ufs_weather_model on Gaea-C5

Commit Message:

• UFSWM - Update modulefiles/ufs_gaea.intel.lua to compile ufs_weather_model on Gaea-C5
• • FV3 -
  • atmos_cubed_sphere - move the call to read_da_inc outside the if external_ic/restart logic

Priority:

• Normal

Git Tracking

UFSWM:

• Closes Compile ufs_weather_model on Gaea-C5 #2268
• Closes update cubed sphere gitmodules for perturbation/increments for cold starts #2271

Sub component Pull Requests:

• FV3: Perturbation/Increments for cold starts NOAA-EMC/fv3atm#832
• atmos_cubed_sphere: Perturbation/Increments for cold starts NOAA-GFDL/GFDL_atmos_cubed_sphere#340

UFSWM Blocking Dependencies:

• None

Changes

Regression Test Changes (Please commit test_changes.list):

• No Baseline Changes.

Input data Changes:

• None.

Library Changes/Upgrades:

• No Updates

---

Testing Log:

• RDHPCS
  • Hera
  • Orion
  • Hercules
  • Jet
  • Gaea
  • Derecho
• WCOSS2
  • Dogwood/Cactus
  • Acorn
• CI
• opnReqTest (complete task if unnecessary)

[DavidBurrows-NCO]

Hello..This PR is in support of NOAA-EMC/global-workflow#2535. Please let me know how I should proceed with testing. Since this PR will not affect other platforms, do I need to run RT on Hera/Derecho/Hercules? Thanks for your time!

[jkbk2004]

@DavidBurrows-NCO Develop branch builds and runs ok on gaea-c5. Can you explain why we need the code change of this PR? Any issue porting global workflow to gaea-c5?

[BrianCurtis-NOAA]

@DavidBurrows-NCO Develop branch builds and runs ok on gaea-c5. Can you explain why we need the code change of this PR? Any issue porting global workflow to gaea-c5?

@jkbk2004 Have you looked into @DavidBurrows-NCO changes to see if it works for the UFSWM on Gaea-C5?

[jkbk2004]

@DavidBurrows-NCO Develop branch builds and runs ok on gaea-c5. Can you explain why we need the code change of this PR? Any issue porting global workflow to gaea-c5?

@jkbk2004 Have you looked into @DavidBurrows-NCO changes to see if it works for the UFSWM on Gaea-C5?

[jkbk2004]

Sorry for accidentally closing/re-opening. Anyway, @BrianCurtis-NOAA @DavidBurrows-NCO the gaea modulefile of this pr is a lot clean. Let me throw a run on gaea. @RatkoVasic-NOAA can you take a look at this modulefile update for gaea?

[BrianCurtis-NOAA]

Sorry for accidentally closing/re-opening. Anyway, @BrianCurtis-NOAA @DavidBurrows-NCO the gaea modulefile of this pr is a lot clean. Let me throw a run on gaea. @RatkoVasic-NOAA can you take a look at this modulefile update for gaea?

One concerning things to me was the removal of the stack-python as it might cause issues with the abort_dep_tasks.py script.

[DavidBurrows-NCO]

Good Morning...the previous module file didn't contain a load cmake, so you get the following error upon building:

David.Burrows@gaea56 07:58 ufsm $ ./build.sh 
UFS MODEL DIR: /gpfs/f5/epic/scratch/David.Burrows/ufs_fails_to_build_on_gaea/ufsm
cmake: /opt/cray/pe/gcc/10.3.0/snos/lib64/libstdc++.so.6: version `GLIBCXX_3.4.30' not found (required by cmake)

I added the module load cmake then did some clean up as well. The modulefile for gaea now resembles gfs_utils, ufs_utils, and upp.
@BrianCurtis-NOAA I was thinking the same about python. I see a mix of python, stack-python, and no python in the other module files. I can add it back if necessary.
Thanks everyone!

[RatkoVasic-NOAA]

This changes worked for me. @BrianCurtis-NOAA can you give me example so I can try to get failure with abort_dep_tasks.py

[jkbk2004]

This changes worked for me. @BrianCurtis-NOAA can you give me example so I can try to get failure with abort_dep_tasks.py

Mostly with importing python ecflow package in https://github.com/ufs-community/ufs-weather-model/blob/develop/tests/abort_dep_tasks.py#L3. But it sounds like it runs ok. @RatkoVasic-NOAA a decision point on your side is to specifically control thru stack-python. I think its safe to keep moving with the stack-python. @DavidBurrows-NCO can you put the stack-python line back. So we can move on.

[DavidBurrows-NCO]

@BrianCurtis-NOAA @RatkoVasic-NOAA I added stack-python load back into the module file and tested a build.

[BrianCurtis-NOAA]

This changes worked for me. @BrianCurtis-NOAA can you give me example so I can try to get failure with abort_dep_tasks.py

This concern is just based on rt.sh failing with running ECFlow on Gaea. @RatkoVasic-NOAA are you using ecflow to run the rt.sh suite?

[RatkoVasic-NOAA]

@BrianCurtis-NOAA I didn't run full test suite, just about 10 test cases to see if it brakes compilation or some executions. I used rocoto. I can now try with ecflow same thing.

[RatkoVasic-NOAA]

Selected 10 tests worked using both ecflow and rocoto for me.

[BrianCurtis-NOAA]

Selected 10 tests worked using both ecflow and rocoto for me.

OK Great. @jkbk2004 please run the full rt suite on Gaea-C5 using ecflow and we can then merge this PR with another non-baseline changing PR once that's completed.

[jkbk2004]

Selected 10 tests worked using both ecflow and rocoto for me.

OK Great. @jkbk2004 please run the full rt suite on Gaea-C5 using ecflow and we can then merge this PR with another non-baseline changing PR once that's completed.

Sure!

[jkbk2004]

@DavidBurrows-NCO This rap clm_lake case is crashing with this pr. experiment is at /gpfs/f5/epic/scratch/Jong.Kim/RT_RUNDIRS
/Jong.Kim/FV3_RT/rt_48937/rap_clm_lake_debug_intel/err. Develop runs ok. @SamuelTrahanNOAA just FYI

[DavidBurrows-NCO]

@jkbk2004 I am also getting a failure with that test. Here are some highlights from the output:

 0: PASS: fcstRUN phase 1, n_atmsteps =               11 time is        14.979133
  0:  ---isec,seconds        3600        3600
  0:   gfs diags time since last bucket empty:    1.00000000000000      hrs
  0:  write out restart at n_atmsteps=          11  seconds=        3600
  0:  integration length=  0.9166667
  0: PASS: fcstRUN phase 2, n_atmsteps =               11 time is         0.788611
  0:   aft fcst run output time=        3600 FBcount=           8 na=          12
...
  0:  liq_aero  3.4789200E+10  1.1008354E+07  1.7738830E+09
  0:  ice_aero  1.4927048E+08  1.2824743E-13   3184276.
  0:  sgs_tke   149.9711      9.9780387E-05   1.240300
  0: NOTE from PE     0: Potential error in diag_manager_end: slp NOT available, check if output interval > runlength. Netcdf fill_values are written
  0: NOTE from PE     0: Potential error in diag_manager_end: vort850 NOT available, check if output interval > runlength. Netcdf fill_values are written
  0: NOTE from PE     0: Potential error in diag_manager_end: vort200 NOT available, check if output interval > runlength. Netcdf fill_values are written
  0: NOTE from PE     0: Potential error in diag_manager_end: us NOT available, check if output interval > runlength. Netcdf fill_values are written
  0: NOTE from PE     0: Potential error in diag_manager_end: u1000 NOT available, check if output interval > runlength. Netcdf fill_values are written
...
144: Stack trace terminated abnormally.
145: forrtl: error (73): floating divide by zero

Are you getting the same error? I don't see any other hints in the working directory for a reason for this failure.

[SamuelTrahanNOAA]

I cannot see your directories due to restrictive permissions.

What you've shown me so far are errors that have nothing to do with CLM lake. I cannot debug farther until I see the stack trace.

[DavidBurrows-NCO]

@SamuelTrahanNOAA Can I move the working dir somewhere on Gaea for you? My Gaea groups are epic and ufs-ard.
The only stack trace I see in the working dir is:

144: Stack trace terminated abnormally.
145: forrtl: error (73): floating divide by zero

Thanks, Sam!

[SamuelTrahanNOAA]

I move the working dir somewhere on Gaea for you?

I don't know where your working directory resides.

You don't necessarily need to move it; I just need the permissions to allow world access.

This is not a stack trace:

145: forrtl: error (73): floating divide by zero

The stack trace would include a trace of the files and line numbers to the point in the stack with the divide by zero.

[DavidBurrows-NCO]

@SamuelTrahanNOAA The working dir is here: /gpfs/f5/ufs-ard/scratch/David.Burrows/RT_RUNDIRS/David.Burrows/FV3_RT/rt_200821.
perms for files there are: -rw-r--r--, so you should be good.

[BrianCurtis-NOAA]

@zach1221 @FernandoAndrade-NOAA @BrianCurtis-NOAA Finally, it's all updated ok. I think it's safe to re-do the tests on RDHPCS side.

HYCOM-interface has a change but no .gitmodule entry. Please fix.

[jkbk2004]

@zach1221 @FernandoAndrade-NOAA @BrianCurtis-NOAA Finally, it's all updated ok. I think it's safe to re-do the tests on RDHPCS side.

HYCOM-interface has a change but no .gitmodule entry. Please fix.

HYCOM hash is already committed one on emc/develop. Good to go with hash update.

[zach1221]

@jkbk2004 @FernandoAndrade-NOAA I talked to Brian offline. We're good to skip acorn.

[DavidBurrows-NCO]

Thanks for everyone's help with this PR @jkbk2004 @zach1221 @FernandoAndrade-NOAA @BrianCurtis-NOAA @SamuelTrahanNOAA @RatkoVasic-NOAA @DusanJovic-NOAA @junwang-noaa @weihuang-jedi