[
https://issues.apache.org/jira/browse/KAFKA-9351?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17009035#comment-17009035
]
Ryanne Dolan commented on KAFKA-9351:
-------------------------------------
[~nitishgoyal13] are those event counts or offsets? MM2 cannot guarantee that
downstream offsets or event counts match exactly -- it only offers
at-least-once semantics. So you can be sure that records are not dropped, and
they maintain approximately the same order, but there may be dupes due to
retries in the producer.
I'm working on a PoC for exactly-once semantics, but currently the Connect
framework makes this difficult to get right.
> Higher count in destination cluster using Kafka MM2
> ---------------------------------------------------
>
> Key: KAFKA-9351
> URL: https://issues.apache.org/jira/browse/KAFKA-9351
> Project: Kafka
> Issue Type: Bug
> Components: mirrormaker
> Affects Versions: 2.4.0
> Reporter: Nitish Goyal
> Priority: Blocker
>
> I have setup replication between cluster across different data centres. After
> setting up replication, at times, I am observing higher event count in
> destination cluster
> Below are counts in source and destination cluster
>
> *Source Cluster*
> ```
>
> events_4:0:51048
> events_4:1:52250
> events_4:2:51526
> ```
>
> *Destination Cluster*
> ```
> nm5.events_4:0:53289
> nm5.events_4:1:54569
> nm5.events_4:2:53733
> ```
>
> This is a blocker for us to start using MM2 replicatior
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
