[
https://issues.apache.org/jira/browse/KAFKA-8588?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
jacky updated KAFKA-8588:
-------------------------
Description:
When partition leader disconnected from zk, I think it should stop accepting
requests, because another leader may be elected by kafkaController, if it
continues to process the request, maybe lead to two leaders in the
cluster(brain split), the result will depend on the acks and
min.insync.replicas config, but I think the server should do something base on
the partition replicas config.
I have read the https://issues.apache.org/jira/browse/KAFKA-6173
was:When partition leader disconnected from zk, I think it should stop
accepting requests, because another leader may be elected by kafkaController,
if it continues to process the request, maybe lead to two leaders in the
cluster(brain split), the result will depend on the acks and
min.insync.replicas config, but I think the server should do something base on
the partition replicas config.
> Leader should stop accepting requests when disconnected from ZK
> ---------------------------------------------------------------
>
> Key: KAFKA-8588
> URL: https://issues.apache.org/jira/browse/KAFKA-8588
> Project: Kafka
> Issue Type: Bug
> Components: log
> Affects Versions: 2.1.1
> Reporter: jacky
> Priority: Major
>
> When partition leader disconnected from zk, I think it should stop accepting
> requests, because another leader may be elected by kafkaController, if it
> continues to process the request, maybe lead to two leaders in the
> cluster(brain split), the result will depend on the acks and
> min.insync.replicas config, but I think the server should do something base
> on the partition replicas config.
> I have read the https://issues.apache.org/jira/browse/KAFKA-6173
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
