[
https://issues.apache.org/jira/browse/HBASE-10130?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13863166#comment-13863166
]
Jeffrey Zhong commented on HBASE-10130:
---------------------------------------
{code}
int version1 = ZKUtil.checkExists(zkw, tasknode);
assertTrue("version1=" + version1 + ", version=" + version, version1 >
version);
{code}
The patch fix looks good to me though I'd prefer to remove the above two lines
because
1)the risking condition on getting the original version
2) when the following verification is true:
{code}
assertEquals(tot_mgr_resubmit.get(), 1);
{code}
The znode version is also verified to be bumped up.
> TestSplitLogManager#testTaskResigned fails sometimes
> ----------------------------------------------------
>
> Key: HBASE-10130
> URL: https://issues.apache.org/jira/browse/HBASE-10130
> Project: HBase
> Issue Type: Test
> Reporter: Ted Yu
> Assignee: Ted Yu
> Priority: Minor
> Attachments: 10130-output.txt, 10130-v1.txt
>
>
> The test failed in
> https://builds.apache.org/job/PreCommit-HBASE-Build/8131//testReport
> For testTaskResigned() :
> {code}
> int version = ZKUtil.checkExists(zkw, tasknode);
> // Could be small race here.
> if (tot_mgr_resubmit.get() == 0) waitForCounter(tot_mgr_resubmit, 0, 1,
> to/2);
> {code}
> There was no log similar to the following (corresponding to waitForCounter()
> call above):
> {code}
> 2013-12-10 21:23:54,905 INFO [main] hbase.Waiter(174): Waiting up to [3,200]
> milli-secs(wait.for.ratio=[1])
> {code}
> Meaning, the version (2) retrieved corresponded to resubmitted task. version1
> retrieved same value, leading to assertion failure.
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
