[infinispan-issues] [JBoss JIRA] (ISPN-4549) StateTransferSuppressForMemcacheIT.testRebalanceWithJoinedNodeStop random failures

[Galder Zamarreño (JIRA)]

    [ https://issues.jboss.org/browse/ISPN-4549?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12990049#comment-12990049 ] 

Galder Zamarreño commented on ISPN-4549:
----------------------------------------

Dan, I'm not sure I understand your assessment, possibly due to lack of knowledge of state transfer. The coordinator only starts rebalance once rebalancing is enabled. I mean, yes, once the coordinator's rebalance flag is disabled and the 3rd node is stopped, it waits for the coordinator to not be in a rebalancing mode, and then it checks the number of entries. TBH, I'm not sure it should be checking for these number of entries, it feels to me these checks are a bit too brittle? In any case, this is probably something that you should address since you're the state transfer expert? :)

> StateTransferSuppressForMemcacheIT.testRebalanceWithJoinedNodeStop	random failures
> ----------------------------------------------------------------------------------
>
>                 Key: ISPN-4549
>                 URL: https://issues.jboss.org/browse/ISPN-4549
>             Project: Infinispan
>          Issue Type: Bug
>      Security Level: Public(Everyone can see) 
>          Components: Test Suite - Server
>    Affects Versions: 7.0.0.Alpha5
>            Reporter: Dan Berindei
>            Assignee: Galder Zamarreño
>            Priority: Blocker
>              Labels: testsuite_stability
>             Fix For: 7.0.0.Beta1
>
>
> The test checks that the rebalance is complete by waiting for the {{CommittedViewAsString}} reported by the RpcManager MBean to be {{"DefaultConsistentHash\{numSegments=60, numOwners=2, members=\[node0/" + getCacheManagerName() + ", node1/" + getCacheManagerName() + "\]\}"}} and the {{PendingViewAsString}} to be {{null}}.
> But immediately after the 3rd server is stopped, the coordinator installs a topology with exactly the same string representation. It only starts the rebalance after that.
> So the test could start checking the number of entries before the rebalance actually started.
> Failure in CI: http://ci.infinispan.org/viewLog.html?buildId=9804&tab=buildResultsDiv&buildTypeId=bt8



--
This message was sent by Atlassian JIRA
(v6.2.6#6264)