Skip the ice specific diffusion

[guillaumevernieres]

Companion PR to NOAA-EMC/jcb-gdas#39

[jswhit2]

just ran a test on hera (with this and NOAA-EMC/jcb-gdas#39) and the gdas_marinebmat job now completes

[jswhit2]

oops - spoke too soon. I see this in the log for gdas_marinebmat

FileNotFoundError: [Errno 2] No such file or directory: '/scratch1/NCEPDEV/stmp2/Jeffrey.S.Whitaker/RUNDIRS/C192coupled3dvar_test/gdas.2021032406/gdasmarineanalysis.2021032406/staticb/hz_ice.nc

[JohnSteffen-NOAA]

@jswhit2 That should be expected since we commented that line out, but I'm not sure what the downstream effects will be (if any).

[jswhit2]

This seems to fix it (at least gdas_marinebmat completes without errors)

diff --git a/ush/python/pygfs/task/marine_bmat.py b/ush/python/pygfs/task/marine_bmat.py
index a4a5b4f1..bae1e948 100644
--- a/ush/python/pygfs/task/marine_bmat.py
+++ b/ush/python/pygfs/task/marine_bmat.py
@@ -287,10 +287,10 @@ class MarineBMat(Task):
                                 f"{self.task_config.APREFIX}{diff_type}_ocean.nc")
             diffusion_coeff_list.append([src, dest])

-        src = os.path.join(self.task_config.DATAstaticb, f"hz_ice.nc")
-        dest = os.path.join(self.task_config.COMOUT_ICE_BMATRIX,
-                            f"{self.task_config.APREFIX}hz_ice.nc")
-        diffusion_coeff_list.append([src, dest])
+        #src = os.path.join(self.task_config.DATAstaticb, f"hz_ice.nc")
+        #dest = os.path.join(self.task_config.COMOUT_ICE_BMATRIX,
+        #                    f"{self.task_config.APREFIX}hz_ice.nc")
+        #diffusion_coeff_list.append([src, dest])

         FileHandler({'copy': diffusion_coeff_list}).sync()

[jswhit2]

however, gdas_marineanlinit fails with

Traceback (most recent call last):
  File "/scratch2/BMC/gsienkf/whitaker/global-workflow/scripts/exglobal_marine_analysis_initialize.py", line 24, in <module>
    MarineAnl.initialize()
  File "/scratch2/BMC/gsienkf/whitaker/global-workflow/ush/python/wxflow/logger.py", line 266, in wrapper
    retval = func(*args, **kwargs)
  File "/scratch2/BMC/gsienkf/whitaker/global-workflow/ush/python/pygfs/task/marine_analysis.py", line 119, in initialize
    self._prep_variational_yaml()
  File "/scratch2/BMC/gsienkf/whitaker/global-workflow/ush/python/wxflow/logger.py", line 266, in wrapper
    retval = func(*args, **kwargs)
  File "/scratch2/BMC/gsienkf/whitaker/global-workflow/ush/python/pygfs/task/marine_analysis.py", line 238, in _prep_variational_yaml
    jedi_config = render(jcb_config)
  File "/scratch2/BMC/gsienkf/whitaker/global-workflow/ush/python/jcb/renderer.py", line 268, in render
    return jcb_object.render(algorithm)
  File "/scratch2/BMC/gsienkf/whitaker/global-workflow/ush/python/jcb/renderer.py", line 199, in render
    jcb.abort_if('{{' in jedi_dict_yaml, f'In template_string_jinja2 '
  File "/scratch2/BMC/gsienkf/whitaker/global-workflow/ush/python/jcb/utilities/trapping.py", line 12, in abort_if
    raise ValueError(message)
ValueError: In template_string_jinja2 the output string still contains template directives. # Set the model component to be used in later configuration

[guillaumevernieres]

@jswhit2 , I got "lucky", looks like I had an hz_ice.nc file in staticb ... But yes, the lines you point to need to be removed. The variational step is going to fail now. There's a formatting issue somewhere. The bug fixes are in a gw pr, but I don't know when this will be looked at let alone merged.
I have a variational job in the queue, I'll let you know if it's worth it to run anything or if the memory problem is on hera as well (it wasn't in @JohnSteffen-NOAA 's build).

[guillaumevernieres]

Back in business:

Same code, different compiler, different memory footprint.

[guillaumevernieres]

@jswhit2 , I isolated the needed fixes from my g-w pr here:
https://github.com/guillaumevernieres/global-workflow/tree/bugfix/config-str-int

can you work on C6?

[jswhit2]

@guillaumevernieres can't use c6 until all the builds support it - there's a bunch of PRs from @DavidBurrows-NCO in to make this happen, and as soon as they are merged I'll give it a try

[guillaumevernieres]

I'm closing this. The bug fixes will be included in NOAA-EMC/global-workflow#3034 and #1353