[jira] [Commented] (ZOOKEEPER-1984) testLeaderTimesoutOnNewQuorum is a flakey test

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

    [ https://issues.apache.org/jira/browse/ZOOKEEPER-1984?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14070949#comment-14070949 ] 

Patrick Hunt commented on ZOOKEEPER-1984:
-----------------------------------------

Alex mentioned:

yep, I think what happens is that server 3 is becoming leader and not server 1, so its not completing the reconfig. Let me think about how to solve this...

then later:

Actually if servers 1 and 3 are talking and 3 is elected and not 1, it means that 3 also saw the reconfig. So it should also complete it when it reboots. To debug this I suggest to print out the last seen config in the beginning of leader.lead(). 

Is it possible that writing the .next file to disk fails ?

Alternatively we could just remove this part of the test (attached patch) - the test's goal is to check that the leader times out when it looses a quorum of the new config, and the part of the test that fails now is not needed to check that. There are other tests in ReconfigRecoveryTest that are supposed to check recovery.

> testLeaderTimesoutOnNewQuorum is a flakey test 
> -----------------------------------------------
>
>                 Key: ZOOKEEPER-1984
>                 URL: https://issues.apache.org/jira/browse/ZOOKEEPER-1984
>             Project: ZooKeeper
>          Issue Type: Bug
>          Components: tests
>    Affects Versions: 3.5.0
>            Reporter: Patrick Hunt
>             Fix For: 3.5.0
>
>
> I'm seeing intermittent failures in testLeaderTimesoutOnNewQuorum
>  It's failing both on jdk6 and jdk7. (this is my personal
> jenkins, I haven't see any other failures than this during the past
> few days).
> {noformat}
> junit.framework.AssertionFailedError
> at org.apache.zookeeper.test.ReconfigTest.testServerHasConfig(ReconfigTest.java:127)
> at org.apache.zookeeper.test.ReconfigTest.testLeaderTimesoutOnNewQuorum(ReconfigTest.java:450)
> at org.apache.zookeeper.JUnit4ZKTestRunner$LoggedInvokeMethod.evaluate(JUnit4ZKTestRunner.java:52)
> {noformat}



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