[
https://issues.apache.org/jira/browse/KAFKA-8229?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16817137#comment-16817137
]
ASF GitHub Bot commented on KAFKA-8229:
---------------------------------------
sdreynolds commented on pull request #6579: KAFKA-8229: Only change nextCommit
if and only if now >= to it
URL: https://github.com/apache/kafka/pull/6579
Summary
When using the commitRequest api within sink context, the next commit
timer should change. It is expected that records get commit on the
next commit interval or it is explicitly requested.
This changes the unit test for commit offset to check to ensure the next
commit in milliseconds doesn't change
### Committer Checklist (excluded from commit message)
- [ ] Verify design and implementation
- [ ] Verify test coverage and CI build status
- [ ] Verify documentation (including upgrade notes)
----------------------------------------------------------------
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:
us...@infra.apache.org
> Connect Sink Task updates nextCommit when commitRequest is true
> ---------------------------------------------------------------
>
> Key: KAFKA-8229
> URL: https://issues.apache.org/jira/browse/KAFKA-8229
> Project: Kafka
> Issue Type: Bug
> Reporter: Scott Reynolds
> Priority: Major
>
> Today, when a WorkerSinkTask uses context.requestCommit(), the next call to
> iteration will cause the commit to happen. As part of the commit execution it
> will also change the nextCommit milliseconds.
> This creates some weird behaviors when a SinkTask calls context.requestCommit
> multiple times. In our case, we were calling requestCommit when the number of
> kafka records we processed exceed a threshold. This resulted in the
> nextCommit being several days in the future and caused it to only commit when
> the record threshold was reached.
> We expected the task to commit when the record threshold was reached OR when
> the timer went off.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
