kafka-jira mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Michael Golovanov (JIRA)" <j...@apache.org>
Subject [jira] [Comment Edited] (KAFKA-6537) Duplicate consumers after consumer group rebalance
Date Thu, 08 Feb 2018 07:47:00 GMT

    [ https://issues.apache.org/jira/browse/KAFKA-6537?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16356606#comment-16356606

Michael Golovanov edited comment on KAFKA-6537 at 2/8/18 7:46 AM:

Seems this issue is duplicate of KAFKA-5430

was (Author: mgolovanov):
Seems this issue duplicate of [KAFKA-5430|https://issues.apache.org/jira/browse/KAFKA-5430]

> Duplicate consumers after consumer group rebalance
> --------------------------------------------------
>                 Key: KAFKA-6537
>                 URL: https://issues.apache.org/jira/browse/KAFKA-6537
>             Project: Kafka
>          Issue Type: Bug
>          Components: consumer
>    Affects Versions:
>            Reporter: Michael Golovanov
>            Priority: Critical
> *_Deployment description_*
> Kafka brokers have been deployed on ten (10) nodes. Zookeeper cluster have seven (7)
nodes. Nodes of Kafka brokers shared with Zookeeper nodes on bare metal hosts.
> Broker/Zookeeper hosts OS is Redhat 7 and JVM version is Java 8. Host names are grid48,
grid237 и grid251.
> We have one topic with six (6) patitions. Kafka consumers deployed on three (3) hosts.
Each host have two (2) consumers. All consumers belong to single group.
> *_Error description_*
> After start all consumers Apache Kafka partitions of topic was balanced evenly.
> grid237 owns partitions 0,1 (0 - consumer thread-0, 1, consumer thread-1)
>  grid251 owns partitions 2,3 (2 - consumer thread-0, consumer thread-1)
>  grid48 owns partitions 4,5 (4- consumer thread-0, 5, consumer thread-1)
> After some period of time we see haotic revokes and assigns partitions between brokers
and then all partitions in log assigned to one consumer on one node grid48
> But really all partitions read not only by thread-1 consumer, but thread-0 on grid48.
And all messages from topic partitions was duplicate. Consumer thread-0 try to commit message
offset and get commit error, thread-1 consumer successfully commit offset.

This message was sent by Atlassian JIRA

View raw message