[
https://issues.apache.org/jira/browse/FLINK-11435?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16754658#comment-16754658
]
Avi Levi commented on FLINK-11435:
----------------------------------
[~elevy] as a kafka user you expect that same consumer group will share the
same queue regardless of the client implementation. This is how kafka
[consumer group|https://kafka.apache.org/documentation/#intro_consumers]
behaves, in this aspect, Flink is an exception. Another advantage is that If
stateless jobs with the same consumer group id will share the same offset,
scaling can be achieved with zero downtime, just add another job. This is how
you do it with other applications as well. I understand the Flink implements
the offset management differently but it doesn't comply to the kafka standard
way (according to the
[documentation|https://kafka.apache.org/documentation/#intro_consumers] )
hence, I do consider it an issue/bug.
> Different jobs same consumer group are treated as separate groups
> -----------------------------------------------------------------
>
> Key: FLINK-11435
> URL: https://issues.apache.org/jira/browse/FLINK-11435
> Project: Flink
> Issue Type: Bug
> Components: Kafka Connector
> Affects Versions: 1.7.1
> Environment: kafka consumer
> Reporter: Avi Levi
> Priority: Major
>
> deploying two jobs with the same consumer group id, still treated as
> different consumer groups. This behavior does not comply with kafka
> expectations. Same consumer group ids should be treated as the same group
> this will enable deploying more jobs (specially if they are stateless ) on
> demand and it also how normal consumer groups behave.
> reproduce :
> deploy the same job twice - both jobs consumes the same message although they
> share the same consumer id
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
