Hello Bo, Doing 1 should be good enough under normal operations. Did you see any errors/exceptions on the controller log regarding partition reassignment?
Guozhang On Mon, Jan 13, 2014 at 6:18 AM, Yu Bo(技术中心) <b...@sohu-inc.com> wrote: > Hi all > Because of badversion exception, the id 1 broker is not a leader. But the > kafka thread on that server is already in work. The id 1 broker is only for > replication. > I want to recover the id 1 broker and became a leader. > I had done : > > 1、 use the “kafka-reassign-partitions.sh” to remove the broker id 1 from > replicas group. And the broker id 1 should be out of touch with kafka > cluster. > > But, the server.log of other servers are printing lots of error,like: > > [2014-01-13 21:56:58,360] INFO Reconnect due to socket error: null > (kafka.consumer.SimpleConsumer) > > [2014-01-13 21:56:58,362] WARN [ReplicaFetcherThread-0-1], Error in fetch > Name: FetchRequest; Version: 0; CorrelationId: 35321893; ClientId: > ReplicaFetcherThread-0-1; ReplicaId: 2; MaxWait: 500 ms; MinBytes: 1 bytes; > RequestInfo: [co,4] -> PartitionFetchInfo(0,1048576) > (kafka.server.ReplicaFetcherThread) > > java.net.ConnectException: Connection refused > > at sun.nio.ch.Net.connect(Native Method) > > at sun.nio.ch.SocketChannelImpl.connect(SocketChannelImpl.java:500) > > at kafka.network.BlockingChannel.connect(BlockingChannel.scala:57) > > at kafka.consumer.SimpleConsumer.connect(SimpleConsumer.scala:44) > > at kafka.consumer.SimpleConsumer.reconnect(SimpleConsumer.scala:57) > > at > kafka.consumer.SimpleConsumer.liftedTree1$1(SimpleConsumer.scala:79) > > at > kafka.consumer.SimpleConsumer.kafka$consumer$SimpleConsumer$$sendRequest(SimpleConsumer.scala:71) > > at > kafka.consumer.SimpleConsumer$$anonfun$fetch$1$$anonfun$apply$mcV$sp$1.apply$mcV$sp(SimpleConsumer.scala:110) > > > 2、kill the kafka thread running on id 1 broker. > > 3、I start the kafka service on broker id 1, and it never add to > replication group. Not to speak of becoming a leader. > > The question is : what should we do inorder to recover the kafka cluster > to original? > > Thanks. > > > > Best regards, > > ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- > 于博 | 搜狐媒体大厦 精准广告研发中心 技术组 | tel:61135569 15101511427 (H) | > b...@sohu-inc.com<mailto:b...@sohu-inc.com> > > -- -- Guozhang