Whitelist generates empty part files sometimes

wri / gfw_forest_loss_geotrellis

Global Tree Cover Loss Analysis using Geotrellis and SPARK

MIT License

10 stars 8 forks source link

Whitelist generates empty part files sometimes #52

Open jterry64 opened 4 years ago

jterry64 commented 4 years ago

Whitelist is occasionally generating empty part files. When trying to upload to GFW API, this causes an empty file error that might put the dataset into a failed state. Easy to workaround, but seems better to just stop it from happening in the first place.

jterry64 commented 4 years ago

Actually, seems like this can just happen occasionally when repartitioning with Spark. There are ways to get rid of empty partitions, but seems like it can happen randomly, so I think I'll just make the datapump more resilient to this happening.