Github user jose-torres commented on the issue:
https://github.com/apache/spark/pull/21490
The proposed strategy still won't work correctly, though. If a batch fails
before commit, the restart won't replay the same records that were in that
batch, but the new ones incoming from the socket.
I agree that the current state is non-ideal, but IMO we should either make
batches actually replayable or add an exception explicitly stating that it's
not allowed to restart a text socket source.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
