You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@kafka.apache.org by "Guozhang Wang (JIRA)" <ji...@apache.org> on 2014/04/15 23:43:16 UTC

[jira] [Commented] (KAFKA-1396) fix unit test ProducerFailureHandlingTest.testBrokerFailure

    [ https://issues.apache.org/jira/browse/KAFKA-1396?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13970090#comment-13970090 ] 

Guozhang Wang commented on KAFKA-1396:
--------------------------------------

1. I tried the test several times but cannot re-produce the hanging locally. So I can only debug based on the stack trace in the comments.

2. Following the stack trace, there is no "synchronized" function call along the trace; also it is not clear if a real dead-lock happens during the hanging.

3. It is suggested that the JVM GC could block any threads:

http://stackoverflow.com/questions/7067058/java-thread-dump-blocked-thread-without-waiting-to-lock
http://stackoverflow.com/questions/4016356/java-blocking-issue-why-would-jvm-block-threads-in-many-different-classes-metho

I agree this test takes long, but it is not a unit test but rather an integration test. I am wondering if we are sure there is occasional forever-hangs there or just occasional taking-too-long-due-to-GC? 

> fix unit test ProducerFailureHandlingTest.testBrokerFailure
> -----------------------------------------------------------
>
>                 Key: KAFKA-1396
>                 URL: https://issues.apache.org/jira/browse/KAFKA-1396
>             Project: Kafka
>          Issue Type: Bug
>          Components: core
>    Affects Versions: 0.8.2
>            Reporter: Jun Rao
>            Assignee: Guozhang Wang
>
> Currently disabled after kafka-1390.



--
This message was sent by Atlassian JIRA
(v6.2#6252)