kafka-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "sy.pan" <shengyi....@gmail.com>
Subject Re: How to recover from out of sync
Date Fri, 07 Aug 2015 03:18:32 GMT
Hi, guys:
	I have fixed the problem just restart kafka broker 0…

	According to http://qnalist.com/questions/4968010/isr-not-updating <http://qnalist.com/questions/4968010/isr-not-updating>
	
>>> Auto leader balancing has some know issues when using together with controlled
shutdown. So, we don't >>> recommend it to be turned on in 0.8.1.1
>>> Thanks,
>>> Jun


	So i set 

	auto.leader.rebalance.enable=false ,
	controlled.shutdown.enable=true

	And then restart kafka, the replicas all in isr. I don’t know why, could someone explain
this ?

	Thank you




> 在 2015年8月6日,18:47,sy.pan <shengyi.pan@gmail.com> 写道:
> 
> Hi guys:
>     
>     we have used kafka-0.8.1 under three machines (broker 0,broker 1,broker 3). After
running several months, some partions is out of sync; how to recover from this situation ?
> 
> topic desc:
> 
> Topic:analyze	PartitionCount:8	ReplicationFactor:3	Configs:
> 	Topic: analyze	Partition: 0	Leader: 1	Replicas: 1,2,0	Isr: 1,2
> 	Topic: analyze	Partition: 1	Leader: 2	Replicas: 2,0,1	Isr: 2,1
> 	Topic: analyze	Partition: 2	Leader: 0	Replicas: 0,1,2	Isr: 1,2,0
> 	Topic: analyze	Partition: 3	Leader: 1	Replicas: 1,0,2	Isr: 1,2
> 	Topic: analyze	Partition: 4	Leader: 2	Replicas: 2,1,0	Isr: 2,1
> 	Topic: analyze	Partition: 5	Leader: 0	Replicas: 0,2,1	Isr: 2,1,0
> 	Topic: analyze	Partition: 6	Leader: 1	Replicas: 1,2,0	Isr: 1,2
> 	Topic: analyze	Partition: 7	Leader: 2	Replicas: 2,0,1	Isr: 2,1
> 
> zk info:
> 
> [zk: localhost:2181(CONNECTED) 2] get /brokers/topics/analyze/partitions/0/state
> {"controller_epoch":1,"leader":1,"version":1,"leader_epoch":1,"isr":[1,2]}
> 
> Here is the related log in borker 0
> 
> state-change.log
> 
> ./state-change.log.2015-08-01:[2015-08-01 06:01:51,449] WARN Broker 0 received invalid
LeaderAndIsr request with correlation id 5 from controller 2 epoch 2 with an older leader
epoch 0 for partition [analyze,0], current leader epoch is 0 (state.change.logger)
> 
> 
> controller.log
> 
> ./controller.log:[2015-08-01 06:01:51,395] WARN [Channel manager on controller 0]: Not
sending request Name: StopReplicaRequest; Version: 0; CorrelationId: 39; ClientId: ; DeletePartitions:
false; ControllerId: 0; ControllerEpoch: 1; Partitions: [analyze,0] to broker 0, since it
is offline. (kafka.controller.ControllerChannelManager)
> 
> 
> Thanks and regards
> 


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message