[
https://issues.apache.org/jira/browse/CAMEL-18350?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17609931#comment-17609931
]
Espen Andreassen commented on CAMEL-18350:
------------------------------------------
I think this bug is somewhat critical. It renders the only technique I know of
for approximating "exactly once" useless. As stated in the description, it will
cause redelivery of all records on topics in case of an error, which probably
is pretty critical for users who have applied an exactly once technique.
Here is a post on stackoverflow demonstrating that this technique probably is
used by more than me:
[https://stackoverflow.com/questions/61621755/how-to-implement-exactly-once-kafka-consumer-using-apache-kafka]
> camel-kafka: combination of breakOnFirstError and maxPollRecords causes camel
> to reconsume all records on error
> ----------------------------------------------------------------------------------------------------------------
>
> Key: CAMEL-18350
> URL: https://issues.apache.org/jira/browse/CAMEL-18350
> Project: Camel
> Issue Type: Bug
> Components: camel-kafka
> Affects Versions: 3.18.0
> Reporter: Espen Andreassen
> Priority: Minor
> Fix For: 3.20.0
>
>
> *Summary:*
> When configuring the camel consumer with props "breakOnFirstError" = "true"
> and "maxPollRecords" = "1", camel reconsumes all messages on topic on error.
> Camel should only have reconsumed the failed message is such cases.
> I configure the component like this to approximate an "exactly once" delivery
> policy (in comibnation with manual commits).
> I've created a project on github with a failing test that recreates the
> issue:
> [https://github.com/espeandr/camel-kafka-incorrectly-reconsumes-entire-topic-demo]
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
