Home | About | Sematext search-lucene.com search-hadoop.com
 Search Hadoop and all its subprojects:

Switch to Threaded View
Kafka >> mail # user >> Leader not local for partition error?


Copy link to this message
-
Leader not local for partition error?
Hi all,

I'm not sure if I'm doing something wrong or if I missed a step
somewhere. A little while ago I successfully got the 0.8 quickstart
example to work fine with the console producer/consumer. Then I went to
work on some code to learn how to implement a producer, which failed
with the producer not being able to send anything with the following
error in the logs:

  Produce request with correlation id 11 failed due to [test,0]: kafka.common.NotLeaderForPartitionException

So I went back to trying the console producer, and I'm getting the same
error. To be sure, I removed all generated data by ZooKeeper and Kafka
and re-followed the steps of the quickstart guide, but I'm getting the
same error with the console producer/consumer.

kafka-list-topic.sh correctly lists my 1-partition, 1-replica test
topic:

  topic: test partition: 0  leader: 0 replicas: 0 isr: 0

ZK and the broker are of course both up and running. Starting the
producer nothing out of the ordinary happens, but when starting the
consumer (before attempting to send anything), I get the following
exception:

  [2013-07-15 13:25:46,487] INFO [ConsumerFetcherThread-console-consumer-943_polygon-1373919946074-f478ba53-0-0], Starting  (kafka.consumer.ConsumerFetcherThread)
  [2013-07-15 13:25:46,517] ERROR [console-consumer-943_polygon-1373919946074-f478ba53-leader-finder-thread], Error due to  (kafka.consumer.ConsumerFetcherManager$LeaderFinderThread)
  kafka.common.NotLeaderForPartitionException
    at sun.reflect.NativeConstructorAccessorImpl.newInstance0(Native Method)

I'm at a loss at what's going on here. When the broker starts it clearly
goes through the election process and becomes the leader (since it's the
only broker anyway) for the 'test' topic:

  [2013-07-15 13:33:05,345] INFO 0 successfully elected as leader (kafka.server.ZookeeperLeaderElector)
  [2013-07-15 13:33:05,550] INFO [Replica Manager on Broker 0]: Handling LeaderAndIsr request Name:LeaderAndIsrRequest;Version:0;Controller:0;ControllerEpoch:3;CorrelationId:0;ClientId:id_0-host_null-port_9092;PartitionState:(test,0) -> (LeaderAndIsrInfo:(Leader:0,ISR:0,LeaderEpoch:0,ControllerEpoch:1),ReplicationFactor:1),AllReplicas:0);Leaders:id:0,host:polygon.turn.com,port:9092 (kafka.server.ReplicaManager)
  [2013-07-15 13:33:05,551] INFO New leader is 0 (kafka.server.ZookeeperLeaderElector$LeaderChangeListener)
  [2013-07-15 13:33:05,563] INFO [Kafka Server 0], Started (kafka.server.KafkaServer)
  [2013-07-15 13:33:05,563] INFO [ReplicaFetcherManager on broker 0] Removing fetcher for partition [test,0] (kafka.server.ReplicaFetcherManager)
  [2013-07-15 13:33:05,566] INFO [Replica Manager on Broker 0]: Handled leader and isr request Name:LeaderAndIsrRequest;Version:0;Controller:0;ControllerEpoch:3;CorrelationId:0;ClientId:id_0-host_null-port_9092;PartitionState:(test,0) -> (LeaderAndIsrInfo:(Leader:0,ISR:0,LeaderEpoch:0,ControllerEpoch:1),ReplicationFactor:1),AllReplicas:0);Leaders:id:0,host:polygon.turn.com,port:9092 (kafka.server.ReplicaManager)

I'm running Kafka from branch 0.8 (b1891e7). Any idea what's going on there?

Thanks,
/Max

--
Maxime Petazzoni
Sr. Platform Engineer
m 408.310.0595
www.turn.com