You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@kafka.apache.org by BigData dev <bi...@gmail.com> on 2017/06/08 22:27:36 UTC
Info regarding kafka topic
Hi,
I have a 3 node Kafka Broker cluster.
I have created a topic and the leader for the topic is broker 1(1001). And
the broker got died.
But when I see the information in zookeeper for the topic, I see the leader
is still set to broker 1 (1001) and isr is set to 1001. Is this a bug in
kafka, as now leader is died, the leader should have set to none.
*[zk: localhost:2181(CONNECTED) 7] get
/brokers/topics/t3/partitions/0/state*
*{"controller_epoch":1,"leader":1001,"version":1,"leader_epoch":1,"isr":[1001]}*
*cZxid = 0x100000078*
*ctime = Thu Jun 08 14:50:07 PDT 2017*
*mZxid = 0x10000008c*
*mtime = Thu Jun 08 14:51:09 PDT 2017*
*pZxid = 0x100000078*
*cversion = 0*
*dataVersion = 1*
*aclVersion = 0*
*ephemeralOwner = 0x0*
*dataLength = 78*
*numChildren = 0*
*[zk: localhost:2181(CONNECTED) 8] *
And when I use describe command the output is
*[root@meets2 kafka-broker]# bin/kafka-topics.sh --describe --topic t3
--zookeeper localhost:2181*
*Topic:t3 PartitionCount:1 ReplicationFactor:2 Configs:*
*Topic: t3 Partition: 0 Leader: 1001 Replicas: 1001,1003 Isr: 1001*
When I use unavailable-partition option, I can know correctly.
*[root@meets2 kafka-broker]# bin/kafka-topics.sh --describe --topic t3
--zookeeper localhost:2181 --unavailable-partitions*
* Topic: t3 Partition: 0 Leader: 1001 Replicas: 1001,1003 Isr: 1001*
But in zookeeper topic state, the leader should have been set to none, not
the actual leader when the broker has died. Is this according to design or
is it a bug in Kafka. Could you please provide any information on this?
*Thanks,*
*Bharat*