Why does Dataflow erratically fail in Datastore access?
问题 My simple Dataflow pipeline successfully copies multiple Kinds from one project's Datastore to another in most cases. But in certain Kinds (about 5% of them), we always get these errors. Dataflow retries 4-8 times with a delay of about 75 seconds, and then the pipeline fails. How can I diagnose and resolve this? EDIT: The answer includes: (1) there was a bug in the Datastore library used by Dataflow; after they fixed this bug, you can see the underlying cause and (2) the default batch size