You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hbase.apache.org by "Lars Hofhansl (JIRA)" <ji...@apache.org> on 2012/11/15 05:40:12 UTC

[jira] [Created] (HBASE-7166) TestSplitTransactionOnCluster tests are flaky

Lars Hofhansl created HBASE-7166:
------------------------------------

             Summary: TestSplitTransactionOnCluster tests are flaky
                 Key: HBASE-7166
                 URL: https://issues.apache.org/jira/browse/HBASE-7166
             Project: HBase
          Issue Type: Bug
            Reporter: Lars Hofhansl
             Fix For: 0.96.0, 0.94.4


There's a variety of tests in this class that fail occasionally.
I think this is caused by incorrect waiting for the split to finish.

The local split method in the test does not wait until both daughters are online, and in some tests there's an assert following immediately that the two daughters exist.


--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

[jira] [Updated] (HBASE-7166) TestSplitTransactionOnCluster tests are flaky

Posted by "Lars Hofhansl (JIRA)" <ji...@apache.org>.
     [ https://issues.apache.org/jira/browse/HBASE-7166?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Lars Hofhansl updated HBASE-7166:
---------------------------------

    Attachment: 7166-0.94-v3.txt

*v3 is what I committed.
Hopefully this will decrease the failure rate of these tests.
                
> TestSplitTransactionOnCluster tests are flaky
> ---------------------------------------------
>
>                 Key: HBASE-7166
>                 URL: https://issues.apache.org/jira/browse/HBASE-7166
>             Project: HBase
>          Issue Type: Bug
>            Reporter: Lars Hofhansl
>            Assignee: Lars Hofhansl
>             Fix For: 0.96.0, 0.94.4
>
>         Attachments: 7166-0.94.txt, 7166-0.94-v2.txt, 7166-0.94-v3.txt, 7166-trunk.txt, 7166-trunk.txt, 7166-trunk-v2.txt, 7166-trunk-v3.txt
>
>
> There's a variety of tests in this class that fail occasionally.
> I think this is caused by incorrect waiting for the split to finish.
> The local split method in the test does not wait until both daughters are online, and in some tests there's an assert following immediately that the two daughters exist.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

[jira] [Commented] (HBASE-7166) TestSplitTransactionOnCluster tests are flaky

Posted by "Ted Yu (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/HBASE-7166?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13498290#comment-13498290 ] 

Ted Yu commented on HBASE-7166:
-------------------------------

{code}
+        Thread.sleep(1000);
{code}
Would suggest shortening the sleep interval. You can increase number of iterations if 10 second wait is needed.
                
> TestSplitTransactionOnCluster tests are flaky
> ---------------------------------------------
>
>                 Key: HBASE-7166
>                 URL: https://issues.apache.org/jira/browse/HBASE-7166
>             Project: HBase
>          Issue Type: Bug
>            Reporter: Lars Hofhansl
>            Assignee: Lars Hofhansl
>             Fix For: 0.96.0, 0.94.4
>
>         Attachments: 7166-0.94.txt, 7166-0.94-v2.txt, 7166-trunk.txt, 7166-trunk.txt, 7166-trunk-v2.txt
>
>
> There's a variety of tests in this class that fail occasionally.
> I think this is caused by incorrect waiting for the split to finish.
> The local split method in the test does not wait until both daughters are online, and in some tests there's an assert following immediately that the two daughters exist.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

[jira] [Commented] (HBASE-7166) TestSplitTransactionOnCluster tests are flaky

Posted by "stack (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/HBASE-7166?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13498596#comment-13498596 ] 

stack commented on HBASE-7166:
------------------------------

/me sweet!
                
> TestSplitTransactionOnCluster tests are flaky
> ---------------------------------------------
>
>                 Key: HBASE-7166
>                 URL: https://issues.apache.org/jira/browse/HBASE-7166
>             Project: HBase
>          Issue Type: Bug
>            Reporter: Lars Hofhansl
>            Assignee: Lars Hofhansl
>             Fix For: 0.96.0, 0.94.4
>
>         Attachments: 7166-0.94.txt, 7166-0.94-v2.txt, 7166-0.94-v3.txt, 7166-trunk.txt, 7166-trunk.txt, 7166-trunk-v2.txt, 7166-trunk-v3.txt
>
>
> There's a variety of tests in this class that fail occasionally.
> I think this is caused by incorrect waiting for the split to finish.
> The local split method in the test does not wait until both daughters are online, and in some tests there's an assert following immediately that the two daughters exist.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

[jira] [Commented] (HBASE-7166) TestSplitTransactionOnCluster tests are flaky

Posted by "Lars Hofhansl (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/HBASE-7166?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13497761#comment-13497761 ] 

Lars Hofhansl commented on HBASE-7166:
--------------------------------------

There appear to be different attempts to fix this. Some in trunk only, and slightly different in different tests.
                
> TestSplitTransactionOnCluster tests are flaky
> ---------------------------------------------
>
>                 Key: HBASE-7166
>                 URL: https://issues.apache.org/jira/browse/HBASE-7166
>             Project: HBase
>          Issue Type: Bug
>            Reporter: Lars Hofhansl
>             Fix For: 0.96.0, 0.94.4
>
>
> There's a variety of tests in this class that fail occasionally.
> I think this is caused by incorrect waiting for the split to finish.
> The local split method in the test does not wait until both daughters are online, and in some tests there's an assert following immediately that the two daughters exist.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

[jira] [Updated] (HBASE-7166) TestSplitTransactionOnCluster tests are flaky

Posted by "Lars Hofhansl (JIRA)" <ji...@apache.org>.
     [ https://issues.apache.org/jira/browse/HBASE-7166?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Lars Hofhansl updated HBASE-7166:
---------------------------------

    Attachment: 7166-0.94-v2.txt

Same for 0.94
                
> TestSplitTransactionOnCluster tests are flaky
> ---------------------------------------------
>
>                 Key: HBASE-7166
>                 URL: https://issues.apache.org/jira/browse/HBASE-7166
>             Project: HBase
>          Issue Type: Bug
>            Reporter: Lars Hofhansl
>            Assignee: Lars Hofhansl
>             Fix For: 0.96.0, 0.94.4
>
>         Attachments: 7166-0.94.txt, 7166-0.94-v2.txt, 7166-trunk.txt, 7166-trunk.txt, 7166-trunk-v2.txt
>
>
> There's a variety of tests in this class that fail occasionally.
> I think this is caused by incorrect waiting for the split to finish.
> The local split method in the test does not wait until both daughters are online, and in some tests there's an assert following immediately that the two daughters exist.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

[jira] [Commented] (HBASE-7166) TestSplitTransactionOnCluster tests are flaky

Posted by "Lars Hofhansl (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/HBASE-7166?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13497790#comment-13497790 ] 

Lars Hofhansl commented on HBASE-7166:
--------------------------------------

Is this test known to have problems with (Open)Jdk7? It seems I cannot run this test at all at home in 0.94.

                
> TestSplitTransactionOnCluster tests are flaky
> ---------------------------------------------
>
>                 Key: HBASE-7166
>                 URL: https://issues.apache.org/jira/browse/HBASE-7166
>             Project: HBase
>          Issue Type: Bug
>            Reporter: Lars Hofhansl
>             Fix For: 0.96.0, 0.94.4
>
>
> There's a variety of tests in this class that fail occasionally.
> I think this is caused by incorrect waiting for the split to finish.
> The local split method in the test does not wait until both daughters are online, and in some tests there's an assert following immediately that the two daughters exist.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

[jira] [Assigned] (HBASE-7166) TestSplitTransactionOnCluster tests are flaky

Posted by "Lars Hofhansl (JIRA)" <ji...@apache.org>.
     [ https://issues.apache.org/jira/browse/HBASE-7166?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Lars Hofhansl reassigned HBASE-7166:
------------------------------------

    Assignee: Lars Hofhansl
    
> TestSplitTransactionOnCluster tests are flaky
> ---------------------------------------------
>
>                 Key: HBASE-7166
>                 URL: https://issues.apache.org/jira/browse/HBASE-7166
>             Project: HBase
>          Issue Type: Bug
>            Reporter: Lars Hofhansl
>            Assignee: Lars Hofhansl
>             Fix For: 0.96.0, 0.94.4
>
>         Attachments: 7166-0.94.txt
>
>
> There's a variety of tests in this class that fail occasionally.
> I think this is caused by incorrect waiting for the split to finish.
> The local split method in the test does not wait until both daughters are online, and in some tests there's an assert following immediately that the two daughters exist.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

[jira] [Commented] (HBASE-7166) TestSplitTransactionOnCluster tests are flaky

Posted by "Ted Yu (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/HBASE-7166?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13498260#comment-13498260 ] 

Ted Yu commented on HBASE-7166:
-------------------------------

{code}
+  private List<HRegion> assertDaughters(byte[] tableName)
{code}
I wouldn't expect an assertion to return a List. Rename the method to getDaughters() ?
                
> TestSplitTransactionOnCluster tests are flaky
> ---------------------------------------------
>
>                 Key: HBASE-7166
>                 URL: https://issues.apache.org/jira/browse/HBASE-7166
>             Project: HBase
>          Issue Type: Bug
>            Reporter: Lars Hofhansl
>            Assignee: Lars Hofhansl
>             Fix For: 0.96.0, 0.94.4
>
>         Attachments: 7166-0.94.txt, 7166-0.94-v2.txt, 7166-trunk.txt, 7166-trunk.txt, 7166-trunk-v2.txt
>
>
> There's a variety of tests in this class that fail occasionally.
> I think this is caused by incorrect waiting for the split to finish.
> The local split method in the test does not wait until both daughters are online, and in some tests there's an assert following immediately that the two daughters exist.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

[jira] [Updated] (HBASE-7166) TestSplitTransactionOnCluster tests are flaky

Posted by "Lars Hofhansl (JIRA)" <ji...@apache.org>.
     [ https://issues.apache.org/jira/browse/HBASE-7166?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Lars Hofhansl updated HBASE-7166:
---------------------------------

    Attachment: 7166-trunk.txt

Oops. Didn't save a file.
                
> TestSplitTransactionOnCluster tests are flaky
> ---------------------------------------------
>
>                 Key: HBASE-7166
>                 URL: https://issues.apache.org/jira/browse/HBASE-7166
>             Project: HBase
>          Issue Type: Bug
>            Reporter: Lars Hofhansl
>            Assignee: Lars Hofhansl
>             Fix For: 0.96.0, 0.94.4
>
>         Attachments: 7166-0.94.txt, 7166-trunk.txt, 7166-trunk.txt
>
>
> There's a variety of tests in this class that fail occasionally.
> I think this is caused by incorrect waiting for the split to finish.
> The local split method in the test does not wait until both daughters are online, and in some tests there's an assert following immediately that the two daughters exist.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

[jira] [Resolved] (HBASE-7166) TestSplitTransactionOnCluster tests are flaky

Posted by "Lars Hofhansl (JIRA)" <ji...@apache.org>.
     [ https://issues.apache.org/jira/browse/HBASE-7166?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Lars Hofhansl resolved HBASE-7166.
----------------------------------

    Resolution: Fixed

Committed to 0.94 and 0.96
                
> TestSplitTransactionOnCluster tests are flaky
> ---------------------------------------------
>
>                 Key: HBASE-7166
>                 URL: https://issues.apache.org/jira/browse/HBASE-7166
>             Project: HBase
>          Issue Type: Bug
>            Reporter: Lars Hofhansl
>            Assignee: Lars Hofhansl
>             Fix For: 0.96.0, 0.94.4
>
>         Attachments: 7166-0.94.txt, 7166-0.94-v2.txt, 7166-0.94-v3.txt, 7166-trunk.txt, 7166-trunk.txt, 7166-trunk-v2.txt, 7166-trunk-v3.txt
>
>
> There's a variety of tests in this class that fail occasionally.
> I think this is caused by incorrect waiting for the split to finish.
> The local split method in the test does not wait until both daughters are online, and in some tests there's an assert following immediately that the two daughters exist.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

[jira] [Commented] (HBASE-7166) TestSplitTransactionOnCluster tests are flaky

Posted by "Hudson (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/HBASE-7166?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13498516#comment-13498516 ] 

Hudson commented on HBASE-7166:
-------------------------------

Integrated in HBase-TRUNK-on-Hadoop-2.0.0 #263 (See [https://builds.apache.org/job/HBase-TRUNK-on-Hadoop-2.0.0/263/])
    HBASE-7166 TestSplitTransactionOnCluster tests are flaky (Revision 1410060)

     Result = SUCCESS
larsh : 
Files : 
* /hbase/trunk/hbase-server/src/test/java/org/apache/hadoop/hbase/regionserver/TestSplitTransactionOnCluster.java

                
> TestSplitTransactionOnCluster tests are flaky
> ---------------------------------------------
>
>                 Key: HBASE-7166
>                 URL: https://issues.apache.org/jira/browse/HBASE-7166
>             Project: HBase
>          Issue Type: Bug
>            Reporter: Lars Hofhansl
>            Assignee: Lars Hofhansl
>             Fix For: 0.96.0, 0.94.4
>
>         Attachments: 7166-0.94.txt, 7166-0.94-v2.txt, 7166-0.94-v3.txt, 7166-trunk.txt, 7166-trunk.txt, 7166-trunk-v2.txt, 7166-trunk-v3.txt
>
>
> There's a variety of tests in this class that fail occasionally.
> I think this is caused by incorrect waiting for the split to finish.
> The local split method in the test does not wait until both daughters are online, and in some tests there's an assert following immediately that the two daughters exist.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

[jira] [Commented] (HBASE-7166) TestSplitTransactionOnCluster tests are flaky

Posted by "Lars Hofhansl (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/HBASE-7166?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13498392#comment-13498392 ] 

Lars Hofhansl commented on HBASE-7166:
--------------------------------------

Fair enough... Going to commit with Ted suggestions. Thanks Ted.
                
> TestSplitTransactionOnCluster tests are flaky
> ---------------------------------------------
>
>                 Key: HBASE-7166
>                 URL: https://issues.apache.org/jira/browse/HBASE-7166
>             Project: HBase
>          Issue Type: Bug
>            Reporter: Lars Hofhansl
>            Assignee: Lars Hofhansl
>             Fix For: 0.96.0, 0.94.4
>
>         Attachments: 7166-0.94.txt, 7166-0.94-v2.txt, 7166-trunk.txt, 7166-trunk.txt, 7166-trunk-v2.txt
>
>
> There's a variety of tests in this class that fail occasionally.
> I think this is caused by incorrect waiting for the split to finish.
> The local split method in the test does not wait until both daughters are online, and in some tests there's an assert following immediately that the two daughters exist.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

[jira] [Updated] (HBASE-7166) TestSplitTransactionOnCluster tests are flaky

Posted by "Lars Hofhansl (JIRA)" <ji...@apache.org>.
     [ https://issues.apache.org/jira/browse/HBASE-7166?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Lars Hofhansl updated HBASE-7166:
---------------------------------

    Attachment: 7166-trunk.txt

Here's a 0.96 patch.
I ran TestSplitTransactionOnCluster locally.
                
> TestSplitTransactionOnCluster tests are flaky
> ---------------------------------------------
>
>                 Key: HBASE-7166
>                 URL: https://issues.apache.org/jira/browse/HBASE-7166
>             Project: HBase
>          Issue Type: Bug
>            Reporter: Lars Hofhansl
>            Assignee: Lars Hofhansl
>             Fix For: 0.96.0, 0.94.4
>
>         Attachments: 7166-0.94.txt, 7166-trunk.txt
>
>
> There's a variety of tests in this class that fail occasionally.
> I think this is caused by incorrect waiting for the split to finish.
> The local split method in the test does not wait until both daughters are online, and in some tests there's an assert following immediately that the two daughters exist.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

[jira] [Commented] (HBASE-7166) TestSplitTransactionOnCluster tests are flaky

Posted by "Hudson (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/HBASE-7166?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13498480#comment-13498480 ] 

Hudson commented on HBASE-7166:
-------------------------------

Integrated in HBase-TRUNK #3548 (See [https://builds.apache.org/job/HBase-TRUNK/3548/])
    HBASE-7166 TestSplitTransactionOnCluster tests are flaky (Revision 1410060)

     Result = SUCCESS
larsh : 
Files : 
* /hbase/trunk/hbase-server/src/test/java/org/apache/hadoop/hbase/regionserver/TestSplitTransactionOnCluster.java

                
> TestSplitTransactionOnCluster tests are flaky
> ---------------------------------------------
>
>                 Key: HBASE-7166
>                 URL: https://issues.apache.org/jira/browse/HBASE-7166
>             Project: HBase
>          Issue Type: Bug
>            Reporter: Lars Hofhansl
>            Assignee: Lars Hofhansl
>             Fix For: 0.96.0, 0.94.4
>
>         Attachments: 7166-0.94.txt, 7166-0.94-v2.txt, 7166-0.94-v3.txt, 7166-trunk.txt, 7166-trunk.txt, 7166-trunk-v2.txt, 7166-trunk-v3.txt
>
>
> There's a variety of tests in this class that fail occasionally.
> I think this is caused by incorrect waiting for the split to finish.
> The local split method in the test does not wait until both daughters are online, and in some tests there's an assert following immediately that the two daughters exist.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

[jira] [Commented] (HBASE-7166) TestSplitTransactionOnCluster tests are flaky

Posted by "Lars Hofhansl (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/HBASE-7166?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13498542#comment-13498542 ] 

Lars Hofhansl commented on HBASE-7166:
--------------------------------------

Yeah baby, the first successful jenkins run in trunk since almost 30 builds!
(0.94 failed on something else :( )
                
> TestSplitTransactionOnCluster tests are flaky
> ---------------------------------------------
>
>                 Key: HBASE-7166
>                 URL: https://issues.apache.org/jira/browse/HBASE-7166
>             Project: HBase
>          Issue Type: Bug
>            Reporter: Lars Hofhansl
>            Assignee: Lars Hofhansl
>             Fix For: 0.96.0, 0.94.4
>
>         Attachments: 7166-0.94.txt, 7166-0.94-v2.txt, 7166-0.94-v3.txt, 7166-trunk.txt, 7166-trunk.txt, 7166-trunk-v2.txt, 7166-trunk-v3.txt
>
>
> There's a variety of tests in this class that fail occasionally.
> I think this is caused by incorrect waiting for the split to finish.
> The local split method in the test does not wait until both daughters are online, and in some tests there's an assert following immediately that the two daughters exist.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

[jira] [Updated] (HBASE-7166) TestSplitTransactionOnCluster tests are flaky

Posted by "Lars Hofhansl (JIRA)" <ji...@apache.org>.
     [ https://issues.apache.org/jira/browse/HBASE-7166?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Lars Hofhansl updated HBASE-7166:
---------------------------------

    Attachment: 7166-trunk-v3.txt
    
> TestSplitTransactionOnCluster tests are flaky
> ---------------------------------------------
>
>                 Key: HBASE-7166
>                 URL: https://issues.apache.org/jira/browse/HBASE-7166
>             Project: HBase
>          Issue Type: Bug
>            Reporter: Lars Hofhansl
>            Assignee: Lars Hofhansl
>             Fix For: 0.96.0, 0.94.4
>
>         Attachments: 7166-0.94.txt, 7166-0.94-v2.txt, 7166-trunk.txt, 7166-trunk.txt, 7166-trunk-v2.txt, 7166-trunk-v3.txt
>
>
> There's a variety of tests in this class that fail occasionally.
> I think this is caused by incorrect waiting for the split to finish.
> The local split method in the test does not wait until both daughters are online, and in some tests there's an assert following immediately that the two daughters exist.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

[jira] [Commented] (HBASE-7166) TestSplitTransactionOnCluster tests are flaky

Posted by "Hudson (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/HBASE-7166?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13498512#comment-13498512 ] 

Hudson commented on HBASE-7166:
-------------------------------

Integrated in HBase-0.94 #595 (See [https://builds.apache.org/job/HBase-0.94/595/])
    HBASE-7166 TestSplitTransactionOnCluster tests are flaky (Revision 1410064)

     Result = FAILURE
larsh : 
Files : 
* /hbase/branches/0.94/src/test/java/org/apache/hadoop/hbase/regionserver/TestSplitTransactionOnCluster.java

                
> TestSplitTransactionOnCluster tests are flaky
> ---------------------------------------------
>
>                 Key: HBASE-7166
>                 URL: https://issues.apache.org/jira/browse/HBASE-7166
>             Project: HBase
>          Issue Type: Bug
>            Reporter: Lars Hofhansl
>            Assignee: Lars Hofhansl
>             Fix For: 0.96.0, 0.94.4
>
>         Attachments: 7166-0.94.txt, 7166-0.94-v2.txt, 7166-0.94-v3.txt, 7166-trunk.txt, 7166-trunk.txt, 7166-trunk-v2.txt, 7166-trunk-v3.txt
>
>
> There's a variety of tests in this class that fail occasionally.
> I think this is caused by incorrect waiting for the split to finish.
> The local split method in the test does not wait until both daughters are online, and in some tests there's an assert following immediately that the two daughters exist.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

[jira] [Updated] (HBASE-7166) TestSplitTransactionOnCluster tests are flaky

Posted by "Lars Hofhansl (JIRA)" <ji...@apache.org>.
     [ https://issues.apache.org/jira/browse/HBASE-7166?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Lars Hofhansl updated HBASE-7166:
---------------------------------

    Attachment: 7166-trunk-v2.txt

This one fixes yet another race where a ZK node does not disappear in time.

I think these changes are uncontroversial. Will commit soon, unless there are objections.
                
> TestSplitTransactionOnCluster tests are flaky
> ---------------------------------------------
>
>                 Key: HBASE-7166
>                 URL: https://issues.apache.org/jira/browse/HBASE-7166
>             Project: HBase
>          Issue Type: Bug
>            Reporter: Lars Hofhansl
>            Assignee: Lars Hofhansl
>             Fix For: 0.96.0, 0.94.4
>
>         Attachments: 7166-0.94.txt, 7166-trunk.txt, 7166-trunk.txt, 7166-trunk-v2.txt
>
>
> There's a variety of tests in this class that fail occasionally.
> I think this is caused by incorrect waiting for the split to finish.
> The local split method in the test does not wait until both daughters are online, and in some tests there's an assert following immediately that the two daughters exist.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

[jira] [Commented] (HBASE-7166) TestSplitTransactionOnCluster tests are flaky

Posted by "Lars Hofhansl (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/HBASE-7166?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13498284#comment-13498284 ] 

Lars Hofhansl commented on HBASE-7166:
--------------------------------------

It does assert in that method, though.
Happy to change the name on commit. Fine otherwise?
                
> TestSplitTransactionOnCluster tests are flaky
> ---------------------------------------------
>
>                 Key: HBASE-7166
>                 URL: https://issues.apache.org/jira/browse/HBASE-7166
>             Project: HBase
>          Issue Type: Bug
>            Reporter: Lars Hofhansl
>            Assignee: Lars Hofhansl
>             Fix For: 0.96.0, 0.94.4
>
>         Attachments: 7166-0.94.txt, 7166-0.94-v2.txt, 7166-trunk.txt, 7166-trunk.txt, 7166-trunk-v2.txt
>
>
> There's a variety of tests in this class that fail occasionally.
> I think this is caused by incorrect waiting for the split to finish.
> The local split method in the test does not wait until both daughters are online, and in some tests there's an assert following immediately that the two daughters exist.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

[jira] [Updated] (HBASE-7166) TestSplitTransactionOnCluster tests are flaky

Posted by "Lars Hofhansl (JIRA)" <ji...@apache.org>.
     [ https://issues.apache.org/jira/browse/HBASE-7166?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Lars Hofhansl updated HBASE-7166:
---------------------------------

    Attachment: 7166-0.94.txt

Here's a 0.94 patch, that just waits a bit for the split to finish. Waits up to 10s, which should be good enough.
(Tested on my work machine with JDK 6)
                
> TestSplitTransactionOnCluster tests are flaky
> ---------------------------------------------
>
>                 Key: HBASE-7166
>                 URL: https://issues.apache.org/jira/browse/HBASE-7166
>             Project: HBase
>          Issue Type: Bug
>            Reporter: Lars Hofhansl
>             Fix For: 0.96.0, 0.94.4
>
>         Attachments: 7166-0.94.txt
>
>
> There's a variety of tests in this class that fail occasionally.
> I think this is caused by incorrect waiting for the split to finish.
> The local split method in the test does not wait until both daughters are online, and in some tests there's an assert following immediately that the two daughters exist.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira