[v2.11.0 release] Recipe failures on DKRZ: `OSError: [Errno -101] NetCDF: HDF error`

[ehogan]

Describe the bug Following ESMValCore v2.11.0rc2 testing, a few recipes are failing with OSError: [Errno -101] NetCDF: HDF error:

• recipe_collins13ipcc.yml

  File "/miniforge3/envs/rc2-env/lib/python3.11/site-packages/iris/fileformats/netcdf/_thread_safe_nc.py", line 384, in __setitem__
  dataset = netCDF4.Dataset(self.path, "r+")
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "src/netCDF4/_netCDF4.pyx", line 2470, in netCDF4._netCDF4.Dataset.__init__
  File "src/netCDF4/_netCDF4.pyx", line 2107, in netCDF4._netCDF4._ensure_nc_success
  OSError: [Errno -101] NetCDF: HDF error: '/esmvaltool_output/recipe_collins13ipcc_20240627_164355/preproc/IAV_calc_tas/tas/CMIP5_CCSM4_Amon_piControl_r1i1p1_tas_0250-1300.nc'

• recipe_ipccwg1ar6ch3_atmosphere.yml

  File "/miniforge3/envs/rc2-env/lib/python3.11/site-packages/iris/fileformats/netcdf/_thread_safe_nc.py", line 384, in __setitem__
  dataset = netCDF4.Dataset(self.path, "r+")
  ^^^^^^^^^^^^^^^^^
  File "src/netCDF4/_netCDF4.pyx", line 2470, in netCDF4._netCDF4.Dataset.__init__
  File "src/netCDF4/_netCDF4.pyx", line 2107, in netCDF4._netCDF4._ensure_nc_success
  OSError: [Errno -101] NetCDF: HDF error: '/esmvaltool_output/recipe_ipccwg1ar6ch3_atmosphere_20240627_170126/preproc/IAV_calc_tas_cmip5/tas/CMIP5_bcc-csm1-1-m_Amon_piControl_r1i1p1_tas_0001-0400.nc'

• recipe_tebaldi21esd.yml

  File "/miniforge3/envs/rc2-env/lib/python3.11/site-packages/iris/fileformats/netcdf/_thread_safe_nc.py", line 384, in __setitem__
  dataset = netCDF4.Dataset(self.path, "r+")
  ^^^^^^^^^^^^^^^^^
  File "src/netCDF4/_netCDF4.pyx", line 2470, in netCDF4._netCDF4.Dataset.__init__
  File "src/netCDF4/_netCDF4.pyx", line 2107, in netCDF4._netCDF4._ensure_nc_success
  OSError: [Errno -101] NetCDF: HDF error: '/esmvaltool_output/recipe_tebaldi21esd_20240627_164355/preproc/fig6d_IAV/tas/CMIP6_MIROC6_Amon_piControl_r1i1p1f1_pr_gn_3200-3500.nc'

• recipe_preprocessor_derive_test.yml

  File "/miniforge3/envs/rc2-env/lib/python3.11/site-packages/iris/fileformats/netcdf/_thread_safe_nc.py", line 384, in __setitem__
  dataset = netCDF4.Dataset(self.path, "r+")
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "src/netCDF4/_netCDF4.pyx", line 2470, in netCDF4._netCDF4.Dataset.__init__
  File "src/netCDF4/_netCDF4.pyx", line 2107, in netCDF4._netCDF4._ensure_nc_success
  OSError: [Errno -101] NetCDF: HDF error: '/esmvaltool_output/recipe_preprocessor_derive_test_20240627_180859/preproc/cmip5/ohc/CMIP5_CCSM4_Omon_historical_r1i1p1_ohc_2004-2005.nc'

The first three were failing during ESMValTool v2.10.0 testing. The last recipe started failing during ESMValCore v2.11.0rc1 testing.

I intend to add these to the list of broken recipes via #3662.

[bouweandela]

@ehogan Have you tried running any of these recipes on a different machine, e.g. Jasmin? If I remember correctly, we previously thought the Levante filesystem caused these issues. I just ran recipe_preprocessor_derive_test.yml (except for the cmip6/toz variable, see #3709) on my laptop and it runs without HDF errors.

[ehogan]

@ehogan Have you tried running any of these recipes on a different machine, e.g. Jasmin? If I remember correctly, we previously thought the Levante filesystem caused these issues. I just ran recipe_preprocessor_derive_test.yml (except for the cmip6/toz variable, see #3709) on my laptop and it runs without HDF errors.

I haven't. I have just set up the other three recipes to run on JASMIN, but I won't have the results until tomorrow (given the timings from the v2.9.0 testing):

• recipe_collins13ipcc.yml (time=7:49:15, mem=272.6)
• recipe_ipccwg1ar6ch3_atmosphere.yml (time=1:42:01, mem=179.0)
• recipe_tebaldi21esd.yml (time=1:50:04, mem=135.0)

[ehogan]

I am struggling to get the first two recipes to run on JASMIN (they keep failing with what appear to be various memory related errors), and the third recipe failed due to missing data, even though I had search_esgf: when_missing set in the ESMValTool user configuration file 😞

[ehogan]

Even though there were memory errors, the recipe_collins13ipcc.yml recipe has just completed on JASMIN 🥳

2024-07-03 14:19:22,220 UTC [33611] INFO    Time for running the recipe was: 7:35:57.567925
2024-07-03 14:19:22,518 UTC [33611] INFO    Maximum memory used (estimate): 128.9 GB
[...]
2024-07-03 14:20:02,500 UTC [33611] INFO    Run was successful

[ehogan]

I am struggling to get the first two recipes to run on JASMIN (they keep failing with what appear to be various memory related errors), and the third recipe failed due to missing data, even though I had search_esgf: when_missing set in the ESMValTool user configuration file 😞

The second recipe also failed due to missing data 😞