Hi Sandeep,
looking into the code, can you please elaborate on how the reading
thread holds the lock for ever? From what I understand from the code the
lock is released after each call to reader.advance(). Therefore the
checkpoint should not be blocked "for ever". Am I missing something?
Would it be possible to recreate the experiments to try and isolate
variables? Right now the 3 cases change both beam and flink versions.
On Tue, Jul 26, 2022 at 11:35 PM Kenneth Knowles wrote:
> Bumping this and adding +John Casey who knows
> about KafkaIO and unbounded sources, though
Bumping this and adding +John Casey who knows about
KafkaIO and unbounded sources, though probably less about the FlinkRunner.
It seems you have isolated it to the Flink translation logic. I'm not sure
who would be the best expert to evaluate if that logic is still OK.
Kenn
On Wed, Jun 29, 2022