Fix time offset issue on ICS with GFS nemsio and netcdf files and add new archive file name on HPSS by chan-hoo · Pull Request #457 · ufs-community/ufs-srweather-app

chan-hoo · 2022-11-05T14:56:23Z

DESCRIPTION OF CHANGES:

The surface climatology file names (gfs.t{hh}z.sfcf{fcst_hr:03d}.[nemsio/netcdf]) are added to file_names of nemsio/fcst and netcdf/fcst in parm/data_locations.yml.

When TIME_OFFSET is off, the make_ics task reads gfs.t12z.atmanl.nemsio(netcdf) and gfs.t12z.sfcanl.nemsio(netcdf)
When TIME_OFFSET is on (for example: 6), the make_ics task reads gfs.t06z.atmf006.nemsio(netcdf) and gfs.t06z.sfcf006.nemsio(netcdf). The flag anl is switched to fcst.
In the current list of 'file_names', gfs.t{hh}z.sfcf{fcst_hr:03d} is missing in fcst of FV3GFS.
However, this gfs.t{hh}z.sfcf{fcst_hr:03d} should be removed for make_lbcs to avoid any unnecessary work with the files. Therefore, the flag ics_or_lbcs is added to ush/retrieve_data.py.

New archive file names of GFS data are added to parm/data_locations.yml.

Old file name (with _prod_): com_gfs_prod_gfs.{yyyymmdd}_{hh}.gfs_pgrb2.tar
New file anme (with _v16.2_): com_gfs_v16.2_gfs.{yyyymmdd}_{hh}.gfs_pgrb2.tar

Type of change

Bug fix (non-breaking change which fixes an issue)
New feature (non-breaking change which adds functionality)
Breaking change (fix or feature that would cause existing functionality to not work as expected)
This change requires a documentation update

TESTS CONDUCTED:

ISSUE:

Fixes issue mentioned in #456

CHECKLIST

My code follows the style guidelines in the Contributor's Guide
I have performed a self-review of my own code using the Code Reviewer's Guide
I have commented my code, particularly in hard-to-understand areas
My changes need updates to the documentation. I have made corresponding changes to the documentation
My changes do not require updates to the documentation (explain).
My changes generate no new warnings
New and existing tests pass with my changes
Any dependent changes have been merged and published

CONTRIBUTORS

@danielabdi-noaa , @christinaholtNOAA

…her-app into bugfix/timeoffset_gfs

chan-hoo · 2022-11-05T15:02:54Z

@christinaholtNOAA @danielabdi-noaa, if you have any better idea, please let me know.

danielabdi-noaa

Looks good to me.

danielabdi-noaa · 2022-11-05T17:41:26Z

+    if cla.external_model == "FV3GFS" and cla.ics_or_lbcs == "LBCS":
+        del file_templates['nemsio']['fcst'][1]
+        del file_templates['netcdf']['fcst'][1]
+


I think you may need to add a similar entry to the "nemsio" format for this to work. Changing the format to nemsio for the new test case you added I get this error:

Traceback (most recent call last): File "/scratch2/BMC/gsd-hpcs/Daniel.Abdi/ufs-srweather-app/ush/retrieve_data.py", line 1012, in <module> main(sys.argv[1:]) File "/scratch2/BMC/gsd-hpcs/Daniel.Abdi/ufs-srweather-app/ush/retrieve_data.py", line 814, in main file_templates = get_file_templates( File "/scratch2/BMC/gsd-hpcs/Daniel.Abdi/ufs-srweather-app/ush/retrieve_data.py", line 296, in get_file_templates del file_templates['nemsio']['fcst'][1] IndexError: list assignment index out of range

I think this is happening because the gfs_file_names anchor is shared for all sources, so when it tries hpss and doesn't succeed it deletes the sfc entry. Then when it tries aws, the entry is already gone. I think you can solve this by not using anchors in the yaml file (bad) or by making a deep copy of the dictionary before deleting the entries.

@danielabdi-noaa, nemsio is not available for the date (08/2022). netcdf and grib2 are only available. Do we need to add any other conditions for this?

@danielabdi-noaa, I've added if-statements to check the availability of netcdf and nemsio for the cycle date in jobs/JREGIONAL_GET_EXTRN_MDL_FILES. If nemsio (or netcdf) is not available, the get_extrn_ics/lbcs will fail with an error message.

@chan-hoo I still think we need a safeguard against multiple deletions of the sfc file -- best would be to find another way that does not delete entries in the data_locations dictionary. If hpss does not have the file for some reason (maybe it is down), and we want to try aws next, it will fail because the sfc entry has been deleted by HPSS. The invalid nemsio date run i did (although wrong) shows what could happen if we can't find the file on hpss but could be available on aws. I added this before the del if statement to make it try aws successfully when it could not find it in hpss

file_templates = deepcopy(file_templates)

@danielabdi-noaa, yes, I agree with you. Can you help me? I've sent an invitation to you.

@chan-hoo No big deal, i've pushed the change i think will avoid multiple deletions of the sfc file. It is not an ideal solution so lets wait for @christinaholtNOAA .

…fs-srweather-app into bugfix/timeoffset_gfs

MichaelLueken

@chan-hoo These changes look good to me. I have built your branch and submitted the new nco_grid_RRFS_CONUS_25km_ics_FV3GFS_lbcs_FV3GFS_timeoffset_suite_GFS_v16 WE2E test on Hera. The new error message when setting a file format that has past the maximum CDATE is a great addition! I will go ahead and launch the Jenkins tests for this work now. Once @christinaholtNOAA has had the opportunity to review this work, I will approve.

chan-hoo · 2022-11-07T18:07:41Z

@MichaelLueken , Thank you!

panll · 2022-11-07T19:03:44Z

Tests passed on Here. It looks good to me!

danielabdi-noaa · 2022-11-07T19:56:54Z

@chan-hoo The MET_verification test is going to fail on cheyenne, but is not your fault. I'm going to have to look at the log files when it finishes to figure out why, but would you be willing to remove that test case from fundamental.cheyenne.gnu with your PR?

chan-hoo · 2022-11-07T20:02:44Z

@danielabdi-noaa, updated.

danielabdi-noaa

Thanks!

MichaelLueken

@chan-hoo With the exception of the MET_verification test on Cheyenne (which your changes wouldn't affect), all of the Jenkins tests have successfully passed.

I would like to note that the new nco_grid_RRFS_CONUS_25km_ics_FV3GFS_lbcs_FV3GFS_timeoffset_suite_GFS_v16 test fails on Cheyenne. It appears as though the test data needs to be staged on that machine. Since the data isn't available, the test will fail. Before this test can be moved to either the fundamental or comprehensive test sets, the new ICs and LBCs will need to be brought over to Cheyenne.

MichaelLueken · 2022-11-07T20:52:49Z

Pending no additional feedback from @christinaholtNOAA, I will merge this PR before 5 pm EST.

christinaholtNOAA · 2022-11-07T20:59:33Z

Actually, I just now saw this. Give me just a little while to take a look?

MichaelLueken · 2022-11-07T21:08:07Z

Actually, I just now saw this. Give me just a little while to take a look?

@christinaholtNOAA No rush. I'll wait until you have approved the PR before merging.

MichaelLueken · 2022-11-07T21:12:37Z

@chan-hoo The MET_verification test is going to fail on cheyenne, but is not your fault. I'm going to have to look at the log files when it finishes to figure out why, but would you be willing to remove that test case from fundamental.cheyenne.gnu with your PR?

@danielabdi-noaa It looks like there is an issue with the metplus installation on Cheyenne:

/glade/p/ral/jntp/MET/MET_releases/10.1.1/bin/grid_stat: error while loading shared libraries: libiomp5.so: cannot open shared object file: No such file or directory

I'll run a test using @natalie-perlin's Cheyenne HPC-stack location and see if this error persists.

danielabdi-noaa · 2022-11-07T21:31:44Z

@MichaelLueken Thanks for looking into the issue. There are some more issues to address with EPIC modulefiles which I've summarized in this issue #458

christinaholtNOAA

@chan-hoo I'm sorry for the late response. These emails didn't immediately grab my attention.

I left a couple of comments below with some concerns and suggestions.

christinaholtNOAA · 2022-11-07T21:23:28Z

+
+    # Remove sfc files from fcst in file_names of FV3GFS for LBCs
+    # sfc files needed in fcst when time_offset is not zero.
+    if cla.external_model == "FV3GFS" and cla.ics_or_lbcs == "LBCS":


Can this just be if "lbcs" and "fcst"? That way we don't tightly couple the tool to specifics of the models we use as inputs?

@christinaholtNOAA, I am afraid that it will cause another unexpected errors with other models. For example, 'GDAS', 'GSMGFS', 'RAP', 'HRRR', and 'NAM' have only one file name for 'fcst'. In these cases, we will have the same index error. For 'GEFS', we should not remove the second array from the file names. We just want to remove the 'sfc' file from the list. 'FV3GFS' only has this issue. What do you think of this?

This logic is still pretty brittle. It assumes that there must be at least 2 and that the 2nd one is the surface file. I think that if we want to protect that assumption, we need a functional test.

I also think it's probably safer to do something like delete the entry if "sfc" is in the name of the file. That type of logic may get us the additional functionality we'd need to use offset GDAS netcdf/nemsio as ICS and expand the logic to "lbcs" and "fcst" instead of FV3GFS-specific logic. In general that says "if we're looking for lbcs from forecasts and a surface file is listed, don't try to get it".

As best I can tell, sfc files only show up for GDAS, FV3GFS, and GSMGFS.

I'd prefer to see something like this:

if cla.ics_or_lbcs == "LBCS": for format in ['netcdf', 'nemsio']: for i, tmpl in enumerate(file_templates.get('format', {}).get('fcst', [])): if "sfc" in tmpl: del file_templates[format]['fcst'][i]

It reduces the assumptions about order and the length of the list provided. It assures that we're only removing surface files, and not atm files accidentally. It also helps us if we want to expand our experiments run with offset GDAS files as ICs -- we can just add the sfc file to the GDAS fcst entry.

@christinaholtNOAA, I tested your script, but sfc file was not removed from the list for LBCs: DEBUG: Looking for files like ['gfs.t{hh}z.atmf{fcst_hr:03d}.nemsio', 'gfs.t{hh}z.sfcf{fcst_hr:03d}.nemsio']. Any suggestion?

I found a typo there: 'format' => format. I am testing it again.

Apologies...I was just writing directly in the text box here and didn't do a full test.

@christinaholtNOAA , the tests were completed successfully!! The part has been replaced with your script. Thank you!

christinaholtNOAA · 2022-11-07T21:29:02Z

 data:
  GSMGFS: compath.py ${envir}/gsmgfs/${gsmgfs_ver}/gsmgfs.${PDY}
-  FV3GFS: compath.py ${envir}/gfs/${gfs_ver}/gfs.${PDY}
+  FV3GFS: compath.py ${envir}/gfs/${gfs_ver}/gfs.${PDY}/${cyc}/atmos


Why not use ${hh} here instead of ${cyc}? This is a bit misleading because this ${cyc} is not necessarily aligned with the cyc set for the RRFS config.

Then, there's no need to rename cyc in the ex-script.

yes, I agree. it has been replaced with 'hh'.

christinaholtNOAA · 2022-11-07T21:29:56Z

 dd=${yyyymmddhh:6:2}
 hh=${yyyymmddhh:8:2}
-
+cyc=${hh}


This is potentially a problem. It resets an NCO required variable. This is the GFS cyc that we're looking for, which is a namespace collision with the existing cyc for RRFS.

chan-hoo · 2022-11-08T11:47:34Z

@christinaholtNOAA, no problem at all. I understand. We are receiving so many emails from SRW :) I've replaced 'cyc' with 'hh'. However, I'd like to hear from you about the if-statement condition as I mentioned above.

MichaelLueken · 2022-11-08T16:43:51Z

@chan-hoo I have gone ahead and relaunched the Jenkins test for Cheyenne with the GNU compiler. All tests successfully passed. Once @christinaholtNOAA has approved, I will go ahead and merge this PR.

chan-hoo added 3 commits November 4, 2022 18:27

fix time offset with gfs

a7fef59

add new archive file name and we2e test

9b9658a

Merge branch 'develop' of https://github.com/ufs-community/ufs-srweat…

7a80c20

…her-app into bugfix/timeoffset_gfs

add ics_or_lbcs to test_retrieve.py

7d6d252

danielabdi-noaa reviewed Nov 5, 2022

View reviewed changes

chan-hoo and others added 4 commits November 5, 2022 22:33

Add check availability of nemsio and netcdf

313a3bb

Copy file_templates dict before modifying it.

37ed64f

remove native from xml for dev_transfer on wcoss2

f5e0918

Merge branch 'bugfix/timeoffset_gfs' of https://github.com/chan-hoo/u…

5c8844e

…fs-srweather-app into bugfix/timeoffset_gfs

MichaelLueken reviewed Nov 7, 2022

View reviewed changes

MichaelLueken added the run_we2e_coverage_tests Run the coverage set of SRW end-to-end tests label Nov 7, 2022

correct data locations on wcoss2

1d38fca

panll approved these changes Nov 7, 2022

View reviewed changes

remove met_verification from fundamental.cheyenne.gnu

a021690

danielabdi-noaa approved these changes Nov 7, 2022

View reviewed changes

MichaelLueken approved these changes Nov 7, 2022

View reviewed changes

christinaholtNOAA reviewed Nov 7, 2022

View reviewed changes

replace cyc to hh in data locations

593cde9

make lbcs condition general

4a3e92d

christinaholtNOAA approved these changes Nov 9, 2022

View reviewed changes

MichaelLueken merged commit 04fd9c6 into ufs-community:develop Nov 9, 2022

MichaelLueken mentioned this pull request Nov 9, 2022

EXTRN_MDL_ICS_TIMEOFFSET_HRS does not work for GFS data (nemsio or netcdf) #456

Closed

chan-hoo deleted the bugfix/timeoffset_gfs branch November 9, 2022 17:28

MatthewPyle-NOAA mentioned this pull request Nov 15, 2022

[develop] bugfix: set PDYext and cycext in get_extrn_ics/lbcs for time-offset #475

Merged

23 tasks

Conversation

chan-hoo commented Nov 5, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

DESCRIPTION OF CHANGES:

Type of change

TESTS CONDUCTED:

ISSUE:

CHECKLIST

CONTRIBUTORS

Uh oh!

chan-hoo commented Nov 5, 2022

Uh oh!

danielabdi-noaa left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

chan-hoo Nov 5, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

danielabdi-noaa Nov 6, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

MichaelLueken left a comment

Choose a reason for hiding this comment

Uh oh!

chan-hoo commented Nov 7, 2022

Uh oh!

panll commented Nov 7, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

danielabdi-noaa commented Nov 7, 2022

Uh oh!

chan-hoo commented Nov 7, 2022

Uh oh!

danielabdi-noaa left a comment

Choose a reason for hiding this comment

Uh oh!

MichaelLueken left a comment

Choose a reason for hiding this comment

Uh oh!

MichaelLueken commented Nov 7, 2022

Uh oh!

christinaholtNOAA commented Nov 7, 2022

Uh oh!

MichaelLueken commented Nov 7, 2022

Uh oh!

MichaelLueken commented Nov 7, 2022

Uh oh!

danielabdi-noaa commented Nov 7, 2022

Uh oh!

christinaholtNOAA left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

chan-hoo commented Nov 5, 2022 •

edited

Loading

chan-hoo Nov 5, 2022 •

edited

Loading

danielabdi-noaa Nov 6, 2022 •

edited

Loading

panll commented Nov 7, 2022 •

edited

Loading