[ https://issues.apache.org/jira/browse/KAFKA-3042?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15233011#comment-15233011 ]
Robert Christ commented on KAFKA-3042: -------------------------------------- Thanks Flavio and Jun for looking at this. We had bumped our zookeeper.session.timeout to 60000 which has allowed us to avoid this problem. In order to reproduce the problem again and get a complete set of logs, we moved the timeout back to the default and started rolling the cluster. After 2 brokers rolled we seem to have run in to a different issue. Broker 1 rolled correctly and rejoined and leadership rebalancing occurred. After broker 2 rolled and came back up it now has an inconsistent view of the metadata. It thinks there are only 300 topics and all the other brokers believe there are 700. We are not sure if this is related in any way. We did not see the controller change (neither broker 1 or 2 were the controller) which is how we had reproduced this problem in the past. Should we file this as a separate issue? We are going to go ahead and try to reproduce the original symptoms. > updateIsr should stop after failed several times due to zkVersion issue > ----------------------------------------------------------------------- > > Key: KAFKA-3042 > URL: https://issues.apache.org/jira/browse/KAFKA-3042 > Project: Kafka > Issue Type: Bug > Affects Versions: 0.8.2.1 > Environment: jdk 1.7 > centos 6.4 > Reporter: Jiahongchao > Attachments: controller.log, server.log.2016-03-23-01, > state-change.log > > > sometimes one broker may repeatly log > "Cached zkVersion 54 not equal to that in zookeeper, skip updating ISR" > I think this is because the broker consider itself as the leader in fact it's > a follower. > So after several failed tries, it need to find out who is the leader -- This message was sent by Atlassian JIRA (v6.3.4#6332)