While testing larger spatial extents for UC8, we bumped into a small number of broken geotiffs on our backend that would crash load_collection.
The culprit files have been identified and are being addressed + a periodic scan of file health will be implemented to alert us of these problems. However, I'm of the opinion that oeop processes shouldn't just crash because of storage issues elsewhere. We do have (some) nodata handling for a reason, so I propose we use the skip_broken_datasets=True flag on dc.load() to smooth over such failures in the future.
While testing larger spatial extents for UC8, we bumped into a small number of broken geotiffs on our backend that would crash load_collection.
The culprit files have been identified and are being addressed + a periodic scan of file health will be implemented to alert us of these problems. However, I'm of the opinion that oeop processes shouldn't just crash because of storage issues elsewhere. We do have (some) nodata handling for a reason, so I propose we use the
skip_broken_datasets=True
flag ondc.load()
to smooth over such failures in the future.