kafka-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Joel Koshy (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (KAFKA-915) System Test - Mirror Maker testcase_5001 failed
Date Thu, 04 Jul 2013 00:45:19 GMT

    [ https://issues.apache.org/jira/browse/KAFKA-915?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13699641#comment-13699641
] 

Joel Koshy commented on KAFKA-915:
----------------------------------

This failure is due to the fact that the leaderAndIsr request has not yet made it to the brokers
until after the mirror maker's rebalance completes. This is related to the issue reported
in KAFKA-956. Previously (before we started caching metadata at the brokers) the partition
information was retrieved directly from zk.

The fix for now would be to use the create topic admin before starting the mirror maker (or
move the producer performance start up to well before the mirror maker startup).

                
> System Test - Mirror Maker testcase_5001 failed
> -----------------------------------------------
>
>                 Key: KAFKA-915
>                 URL: https://issues.apache.org/jira/browse/KAFKA-915
>             Project: Kafka
>          Issue Type: Bug
>            Reporter: John Fung
>            Assignee: Joel Koshy
>            Priority: Critical
>              Labels: kafka-0.8, replication-testing
>         Attachments: testcase_5001_debug_logs.tar.gz
>
>
> This case passes if brokers are set to partition = 1, replicas = 1
> It fails if brokers are set to partition = 5, replicas = 3 (consistently reproducible)
> This test case is set up as shown below.
> 1. Start 2 ZK as a cluster in Source
> 2. Start 2 ZK as a cluster in Target
> 3. Start 3 brokers as a cluster in Source (partition = 1, replicas = 1)
> 4. Start 3 brokers as a cluster in Target (partition = 1, replicas = 1)
> 5. Start 1 MM
> 6. Start ProducerPerformance to send some data
> 7. After Producer is done, start ConsoleConsumer to consume data
> 8. Stop all processes and validate if there is any data loss.
> 9. No failure is introduced to any process in this test
> Attached a tar file which contains the logs and system test output for both cases.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

Mime
View raw message