[jira] [Commented] (HBASE-20465) Fix TestEnableRSGroup flaky

["Balazs Meszaros (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HBASE-20465?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16445843#comment-16445843 ] 

Balazs Meszaros commented on HBASE-20465:
-----------------------------------------

I was not able to reproduce the error locally, but I think the cause of the problem is that master stop and start overlap each other, because we did not wait until master stops.

{code:java}
TEST_UTIL.getMiniHBaseCluster().stopMaster(0);
// ...
TEST_UTIL.getMiniHBaseCluster().startMaster();
TEST_UTIL.getMiniHBaseCluster().waitForActiveAndReadyMaster(60000);
 
assertTrue(TEST_UTIL.getMiniHBaseCluster().getMaster() != null);
{code}

One possible scenario for the error:
1. Previous master is still running {{mt.getMaster().isActiveMaster() && mt.getMaster().isInitialized()}}, then {{waitForActiveAndReadyMaster}} will return.
2. Previous master stops, but new master hasn't started yet.
3. {{getMaster}} will return {{null}}, because no master thread satisfies this condition {{mt.getMaster().isActiveMaster()  && !mt.getMaster().isStopped()}}.



> Fix TestEnableRSGroup flaky
> ---------------------------
>
>                 Key: HBASE-20465
>                 URL: https://issues.apache.org/jira/browse/HBASE-20465
>             Project: HBase
>          Issue Type: Bug
>          Components: rsgroup
>    Affects Versions: 2.0.0
>            Reporter: Balazs Meszaros
>            Assignee: Balazs Meszaros
>            Priority: Major
>             Fix For: 2.0.0
>
>         Attachments: HBASE-20465.master.001.patch
>
>
> Most recent {{TestEnableRSGroup}} tests failed on branch-2.0 according to our flaky dashboard.
> {code}
> java.lang.AssertionError
> 	at org.apache.hadoop.hbase.rsgroup.TestEnableRSGroup.testEnableRSGroup(TestEnableRSGroup.java:80)
> {code}
> TestEnableRSGroup:80:
> {code:java}
> // check if master started successfully
> assertTrue(TEST_UTIL.getMiniHBaseCluster().getMaster() != null);
> {code}



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)