[
https://issues.apache.org/jira/browse/KAFKA-451?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15470408#comment-15470408
]
Ismael Juma commented on KAFKA-451:
-----------------------------------
We have the following code in AbstractFetcherThread:
{code}
warn(s"Error in fetch $fetchRequest", t)
inLock(partitionMapLock) {
partitionsWithError ++= partitionMap.keys
// there is an error occurred while fetching partitions, sleep a
while
partitionMapCond.await(fetchBackOffMs, TimeUnit.MILLISECONDS)
}
{code}
Is that what was intended in this JIRA?
> follower replica may need to backoff the fetching if leader is not ready yet
> ----------------------------------------------------------------------------
>
> Key: KAFKA-451
> URL: https://issues.apache.org/jira/browse/KAFKA-451
> Project: Kafka
> Issue Type: Bug
> Components: core
> Reporter: Jun Rao
> Assignee: Prashanth Menon
> Labels: optimization
> Fix For: 0.10.1.0
>
> Original Estimate: 48h
> Remaining Estimate: 48h
>
> Currently, when a follower starts fetching from a new leader, it just keeps
> sending fetch requests even if the requests fail because the leader is not
> ready yet. We probably should let the follower backoff a bit in this case.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
