All facets for all configured data sources of a project are always required

[schlunma]

Not sure if a bug or feature.

If multiple data sources are defined for a single project, all facets are always required. For example, if two sources for dataset A (with facet a) and dataset B (with facet b) exist, and only dataset B is used in the recipe, facet a would still need to be given in the recipe. This can be very confusing (see also discussion here: #2946 (comment)).

This is particularly relevant for the native6 project, which is supposed to be shared among many different datasets. Another example case is the ERA5 GRIB data source, whose facets are very different to its netcdf counterpart:

ESMValCore/esmvalcore/config/configurations/data-hpc-dkrz.yml

Lines 70 to 83 in 974ff8c

	native6:
	data:
	dkrz:
	type: esmvalcore.io.local.LocalDataSource
	rootpath: /work/bd0854/DATA/ESMValTool2/RAWOBS
	dirname_template: "Tier{tier}/{dataset}/{version}/{frequency}/{short_name}"
	filename_template: "*.nc"
	# ERA5 data in GRIB format:
	# https://docs.dkrz.de/doc/dataservices/finding_and_accessing_data/era_data/index.html#pool-data-era5-file-and-directory-names
	dkrz-era5:
	type: esmvalcore.io.local.LocalDataSource
	rootpath: /pool/data/ERA5
	dirname_template: "{family}/{level}/{type}/{tres}/{grib_id}"
	filename_template: "{family}{level}{typeid}_{tres}_*_{grib_id}.grb"

As a consequence, all recipes we have currently published will fail when this default DKRZ configuration is used.

Solutions to this could be:

• Avoid using different facets in different data sources (that also means no sharing of the native6 project among different datasets)
• Make sure that sensible defaults exist for the differing facets (defined as extra facets)

I will open a PR which comments out the GRIB data source for ERA5 on Levante (this causes recipes to break), but this is only a temporary solution.