[
https://issues.apache.org/jira/browse/KAFKA-1407?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13999664#comment-13999664
]
Krzysztof Ociepa commented on KAFKA-1407:
-----------------------------------------
I have the same issue using the latest 0.8.1.1 release:
[2014-05-16 06:32:15,591] INFO Partition [main_topic,5] on broker 2: Shrinking
ISR for partition [main_topic,5] from 3,2,1 to 2 (kafka.cluster.Partition)
[2014-05-16 06:32:15,633] ERROR Conditional update of path
/brokers/topics/main_topic/partitions/5/state with data
{"controller_epoch":6,"leader":2,"version":1,"leader_epoch":26,"isr":[2]} and
expected version 51 failed due to
org.apache.zookeeper.KeeperException$BadVersionException: KeeperErrorCode =
BadVersion for /brokers/topics/main_topic/partitions/5/state
(kafka.utils.ZkUtils$)
[2014-05-16 06:32:15,643] INFO Partition [main_topic,5] on broker 2: Cached
zkVersion [51] not equal to that in zookeeper, skip updating ISR
(kafka.cluster.Partition)
> Broker can not return to ISR because of BadVersionException
> -----------------------------------------------------------
>
> Key: KAFKA-1407
> URL: https://issues.apache.org/jira/browse/KAFKA-1407
> Project: Kafka
> Issue Type: Bug
> Components: controller
> Affects Versions: 0.8.1
> Reporter: Dmitry Bugaychenko
> Assignee: Neha Narkhede
>
> Each morning we found a broker out of ISR at stuck with log full of messages:
> {code}
> INFO | jvm 1 | 2014/04/21 08:36:21 | [2014-04-21 09:36:21,907] ERROR
> Conditional update of path /brokers/topics/topic2/partitions/1/state with
> data
> {"controller_epoch":46,"leader":2,"version":1,"leader_epoch":38,"isr":[2]}
> and expected version 53 failed due to
> org.apache.zookeeper.KeeperException$BadVersionException: KeeperErrorCode =
> BadVersion for /brokers/topics/topic2/partitions/1/state
> (kafka.utils.ZkUtils$)
> INFO | jvm 1 | 2014/04/21 08:36:21 | [2014-04-21 09:36:21,907] INFO
> Partition [topic2,1] on broker 2: Cached zkVersion [53] not equal to that in
> zookeeper, skip updating ISR (kafka.cluster.Partition)
> {code}
> It seems that it can not recover after short netwrok break down and the only
> way to return it is restart it using kill -9.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
