You are viewing a plain text version of this content. The canonical link for it is here.
Posted to common-dev@hadoop.apache.org by "Arun C Murthy (JIRA)" <ji...@apache.org> on 2007/01/22 05:51:30 UTC

[jira] Commented: (HADOOP-912) TestMiniMRWithDFS fails sporadically

    [ https://issues.apache.org/jira/browse/HADOOP-912?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12466389 ] 

Arun C Murthy commented on HADOOP-912:
--------------------------------------

I suspect this is more of a drawback of the TestMiniMRWithDFS than the MR-framework itself...

MiniMrCluster.waitUntilIdle checks for any 'running' tasks on the TaskTracker... post HADOOP-840 the TaskTracker the cleanup actions are queued and are done offline, which the test-case isn't aware of ; which I suspect leads to a timing issue where the TaskTracker is rightly *idle* and the 'cleanup thread' hasn't kicked in yet...

Should we add a 'sleep' in MiniMRCluster.waitUntilIdle or perhaps another api to check if the job has finished and has been cleaned-up? Thoughts?

> TestMiniMRWithDFS fails sporadically
> ------------------------------------
>
>                 Key: HADOOP-912
>                 URL: https://issues.apache.org/jira/browse/HADOOP-912
>             Project: Hadoop
>          Issue Type: Bug
>          Components: mapred
>    Affects Versions: 0.10.1
>            Reporter: Nigel Daley
>         Assigned To: Arun C Murthy
>
> TestMiniMRWithDFS fails sporadically with the following error:
> junit.framework.AssertionFailedError: Spurious directory task_0001_m_000008_0 found in C:\hudson\workspace\Hadoop-WindowsSmokeTest\trunk\build\test\mapred\local\50068_0
> 	at org.apache.hadoop.mapred.TestMiniMRWithDFS.checkTaskDirectories(TestMiniMRWithDFS.java:128)
> 	at org.apache.hadoop.mapred.TestMiniMRWithDFS.testWithDFS(TestMiniMRWithDFS.java:163)
> This tests checks that all mapred.local.dirs have been cleaned up after running a mapred job on MiniMRCluster.  Effectively, this tests waits for TaskTracker.isIdle() to return true before checking the dirs have been cleaned up.  I believe that HADOOP-639 reordered the cleanup code so that isIdle() will return true before the directory cleanup is complete.

-- 
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators: https://issues.apache.org/jira/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira