[
https://issues.apache.org/jira/browse/ZOOKEEPER-1798?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13811090#comment-13811090
]
Thawan Kooburat commented on ZOOKEEPER-1798:
--------------------------------------------
Just for the record, this test is not known to be flaky in our internal Jenkins
(that test our internal branch).
I am able to repro this on my mac. (Java 1.7.0_15, OSX 10.7.5). When this
happen, it looks txnlog doesn't have any valid content in it. So the zkdb that
we loaded after shutting down the observer never have txn that its znodes to
"data2". I also modified the test to leave the data files around and try to
load it manually after the test fail. The txnlog is loaded successfully with
the right content.
I am thinking that the data flushed to disk by one thread is not visible by the
other thread even after thread.join() is called in between. However, this
really seem unlikely. But I ran the same test in our production host, I cannot
repro the issue (yet)
In Patrick log, this is slightly different. The test failed at line 1105, this
means that the first txn in txnlog is read correctly, but not the second one.
> Fix race condition in testNormalObserverRun
> -------------------------------------------
>
> Key: ZOOKEEPER-1798
> URL: https://issues.apache.org/jira/browse/ZOOKEEPER-1798
> Project: ZooKeeper
> Issue Type: Bug
> Reporter: Flavio Junqueira
> Assignee: Thawan Kooburat
> Priority: Blocker
> Fix For: 3.4.6, 3.5.0
>
> Attachments: TEST-org.apache.zookeeper.server.quorum.Zab1_0Test.txt,
> ZOOKEEPER-1798-b3.4.patch, ZOOKEEPER-1798-b3.4.patch,
> ZOOKEEPER-1798-b3.4.patch, ZOOKEEPER-1798.patch, ZOOKEEPER-1798.patch
>
>
> This is the output messges:
> <noformat>
> Testcase: testNormalObserverRun took 4.221 sec
> FAILED
> expected:<data[2]> but was:<data[1]>
> junit.framework.AssertionFailedError: expected:<data[2]> but was:<data[1]>
> at
> org.apache.zookeeper.server.quorum.Zab1_0Test$8.converseWithObserver(Zab1_0Test.java:1118)
> at
> org.apache.zookeeper.server.quorum.Zab1_0Test.testObserverConversation(Zab1_0Test.java:546)
> at
> org.apache.zookeeper.server.quorum.Zab1_0Test.testNormalObserverRun(Zab1_0Test.java:994)
> <noformat>
--
This message was sent by Atlassian JIRA
(v6.1#6144)
