Hi, and thanks for the great work on xcube-eopf and the EOPF ecosystem.
We are new to the zarr format and currently evaluating Zarr-based workflows via EOPF STAC for use in coastal monitoring. xcube-eopf has been very promising so far. However, we are encountering an issue related to missing or temporarily unavailable Zarr stores.
Some STAC items reference Zarr stores that cannot be opened (e.g., object store errors, timeouts, or “not found”). When accessing these stores manually with xarray (e.g. engine="eopf-zarr"), we can implement a simple fault-tolerant approach:
try:
ds = xr.open_dataset(url, engine="eopf-zarr", chunks={})
except Exception:
passThis allows us to skip problematic scenes and continue building a time series from the remaining ones.
When loading data through xcube-eopf directly from a STAC collection, we are unsure how to achieve similar behavior. Is there a recommended way to skip STAC items whose Zarr stores are unavailable and continue loading the rest of the collection without the process failing?
We have prepared a minimal reproducible example in the following repository: https://github.com/pglez82/tests_zarr/blob/master/minimal_example.ipynb
Hi, and thanks for the great work on xcube-eopf and the EOPF ecosystem.
We are new to the zarr format and currently evaluating Zarr-based workflows via EOPF STAC for use in coastal monitoring. xcube-eopf has been very promising so far. However, we are encountering an issue related to missing or temporarily unavailable Zarr stores.
Some STAC items reference Zarr stores that cannot be opened (e.g., object store errors, timeouts, or “not found”). When accessing these stores manually with xarray (e.g. engine="eopf-zarr"), we can implement a simple fault-tolerant approach:
This allows us to skip problematic scenes and continue building a time series from the remaining ones.
When loading data through xcube-eopf directly from a STAC collection, we are unsure how to achieve similar behavior. Is there a recommended way to skip STAC items whose Zarr stores are unavailable and continue loading the rest of the collection without the process failing?
We have prepared a minimal reproducible example in the following repository: https://github.com/pglez82/tests_zarr/blob/master/minimal_example.ipynb