ChrisSamo632 commented on pull request #4822:
URL: https://github.com/apache/nifi/pull/4822#issuecomment-834187070


   > @ChrisSamo632 It seems to me that the checkpoint at shutdown does not 
work. I have a few uncheckpointed messages that get processed again and again 
when I restart the processor.
   > 
   > I don't really understand how the graceful shutdown works in KCL, but it 
seems to me we have no chance to exit cleanly (I am always seeing ZOMBIE state 
in the record processor's shutdown callback).
   > Could you please double check it?
   > 
   > ...
   
   @turcsanyip I'd noticed ZOMBIE logs a few times and thought it seemed a 
"normal" thing for KCL (from what I could tell, but note I'm hardly a 
Kinesis/KCL expert). Taking a look around on the internet, there are many 
threads about people seeing this kind of behaviour in many different versions 
of the KCL, for example:
   * https://github.com/awslabs/amazon-kinesis-client/issues/108
   * https://github.com/awslabs/amazon-kinesis-client/issues/677
   * https://github.com/amazon-archives/amazon-kinesis-connectors/issues/10
   
   A common thing between these seems to be that the KCL settings likely need 
tweaking depending upon one's setup. With the ability to configure most KCL 
settings via Dynamic Properties on the processor, it's largely open for users 
to figure this out I guess depending upon their setup.
   
   That said, a common theme seems to be the need to increase the 
`failoverTimeMillis` and/or `gracefulShutdownMillis` KCL settings, I've 
therefore added those to the `ConsumeKinesisStream` processor's properties to 
encourage users to investigate/tweak as required.


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
[email protected]


Reply via email to