[
https://issues.apache.org/jira/browse/FLUME-2667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14511477#comment-14511477
]
Gwen Shapira commented on FLUME-2667:
-------------------------------------
I have a major concern here:
I'm not sure we want to support both Scala and Java clients as options. Since
the new producer will work with old Kafka brokers, which producer we use is an
internal implementation detail of KafkaSink. If we are not sure the new
Producer works, we shouldn't have this patch. If we are sure it works, why
support the old one?
The decision to support both complicates the implementation and testing quite a
bit, adding layers of abstraction, factories, etc. I don't think we gain enough
to justify the new maintenance complexity.
Do I miss a good reason to support both producers?
> Add new Kafka Java API in flume-ng-kafka-sink which keep compatibility with
> current API
> ---------------------------------------------------------------------------------------
>
> Key: FLUME-2667
> URL: https://issues.apache.org/jira/browse/FLUME-2667
> Project: Flume
> Issue Type: Improvement
> Components: Sinks+Sources
> Affects Versions: v1.6.0
> Reporter: Frank Yao
> Assignee: Frank Yao
> Priority: Minor
> Attachments: FLUME-2667-0.patch
>
>
> Kafka has released 0.8.2 and meanwhile starts to using new producer API which
> is written in Java originally. Currently, we use javaapi written in scala in
> flume-ng-kafka-sink. I've added new Java API into current sink and re-write
> the tests with mockito.
> As far as I know by reading the source code of kafka 0.8.2, new Java API of
> Consumer would seen be released soon. I think in the very soon, kafka API
> will all change into Java API.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
