[jira] [Commented] (KAFKA-5406) NoNodeException result in rebalance failed

["huxihx (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/KAFKA-5406?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16042077#comment-16042077 ] 

huxihx commented on KAFKA-5406:
-------------------------------

Maybe could estimate the total time period for the network recovery and make sure `rebalance.max.retries` * `rebalance.backoff.ms` is no less than the period. Perhaps some application-level logic is required to handle a long network outage.

> NoNodeException result in rebalance failed
> ------------------------------------------
>
>                 Key: KAFKA-5406
>                 URL: https://issues.apache.org/jira/browse/KAFKA-5406
>             Project: Kafka
>          Issue Type: Bug
>          Components: consumer
>    Affects Versions: 0.8.2.2, 0.10.0.0
>         Environment: windows8.1 centos6.4
>            Reporter: xiaoguy
>            Priority: Critical
>              Labels: easyfix, patch
>         Attachments: log.log
>
>
> hey guys , I got this problem this days,
> because of the network is unstable， consumer rebalance failed after 5 times ,the log shows that zk path /consumers/$(groupIdName)/ids/ is empty,
> consumer seems can't register after network recovered, so i got the kafka source code (0.8.2.2) and found the consumer/ZookeeperConsumerConnector$ZKSessionExpireListener handleNewSession won't call , and handleStateChanged do nothing,
> so i change the code like this ,and it seems works ,  and i checked 0.10.0.0 version, the same problem, is this a bug ? i'm confused , thank you.
>   def handleStateChanged(state: KeeperState) {
>        // do nothing, since zkclient will do reconnect for us.
>          if(state==KeeperState.SyncConnected){
>               handleNewSession()
>          }
>       System.err.println("----------------ZKSessionExpireListener------------ handleStateChanged-----state:"+state+"----"+state.getIntValue)
>       }



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)