[
https://issues.apache.org/jira/browse/KAFKA-6067?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16209482#comment-16209482
]
Antony Stubbs commented on KAFKA-6067:
--------------------------------------
Are you including the leader? If all the followers and the leader have crashed,
then another node cannot run the partition without data loss.
Do you mean if just the followers have crashed, but the leader is still alive,
why not assign another broker to become new followers?
> how to process when all isr crashed
> -----------------------------------
>
> Key: KAFKA-6067
> URL: https://issues.apache.org/jira/browse/KAFKA-6067
> Project: Kafka
> Issue Type: New Feature
> Reporter: haiyangyu
>
> when a topic partition's all isr crashed, the partition is Unavailability,
> why don't like hdfs when a replication in isr crashed and auto move the rep
> to another alive one,
> what's the original intention of this design?
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
