Closed dehora closed 7 years ago
It turns out retry limits weren't honored for stream consumption, just calls like checkpointing.
The patch was tested with a simple loop of 100 processors competing for 1 partition, each processor configured with 2 retry attempts.
This shows how threads aren't cleaned up from before the patch:
This is with the patch and shows threads being cleaned up, eventually leaving just one processor active:
This signals that the number of retries for a processor have been met so that the processor can perform a shutdown and clean up its underlying resources via stopStreaming.
For #284.