[infinispan-issues] [JBoss JIRA] (ISPN-5268) Backup segments may not be removed after failed state transfer

[Dan Berindei (JIRA)]

     [ https://issues.jboss.org/browse/ISPN-5268?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Dan Berindei resolved ISPN-5268.
--------------------------------
    Fix Version/s: 7.0.0.Final
       Resolution: Done


This IllegalArgumentException was breaking the topology update handling:

{noformat}
2014/07/31:11:16:11,110 WARN  [org.infinispan.topology.CacheTopologyControlCommand] (remote-thread-931) ISPN000071: Caught exception when handling command CacheTopologyControlCommand{cache=cache, type=CH_UPDATE, sender=node01, joinInfo=null, topologyId=61, currentCH=DefaultConsistentHash{numSegments=240, numOwners=2, members=[...]}, pendingCH=null, throwable=null, viewId=32}: java.lang.IllegalArgumentException: Node node04 is not a member
    at org.infinispan.distribution.ch.DefaultConsistentHash.getSegmentsForOwner(DefaultConsistentHash.java:87) [infinispan-core-6.1.0.Final-redhat-4.jar:6.1.0.Final-redhat-4]
    at org.infinispan.statetransfer.StateConsumerImpl.onTopologyUpdate(StateConsumerImpl.java:457) [infinispan-core-6.1.0.Final-redhat-4.jar:6.1.0.Final-redhat-4]
    at org.infinispan.statetransfer.StateTransferManagerImpl.doTopologyUpdate(StateTransferManagerImpl.java:199) [infinispan-core-6.1.0.Final-redhat-4.jar:6.1.0.Final-redhat-4]
    at org.infinispan.statetransfer.StateTransferManagerImpl.access$000(StateTransferManagerImpl.java:39) [infinispan-core-6.1.0.Final-redhat-4.jar:6.1.0.Final-redhat-4]
    at org.infinispan.statetransfer.StateTransferManagerImpl$1.updateConsistentHash(StateTransferManagerImpl.java:104) [infinispan-core-6.1.0.Final-redhat-4.jar:6.1.0.Final-redhat-4]
    at org.infinispan.topology.LocalTopologyManagerImpl.handleConsistentHashUpdate(LocalTopologyManagerImpl.java:201) [infinispan-core-6.1.0.Final-redhat-4.jar:6.1.0.Final-redhat-4]
    at org.infinispan.topology.CacheTopologyControlCommand.doPerform(CacheTopologyControlCommand.java:152) [infinispan-core-6.1.0.Final-redhat-4.jar:6.1.0.Final-redhat-4]
    at org.infinispan.topology.CacheTopologyControlCommand.perform(CacheTopologyControlCommand.java:124) [infinispan-core-6.1.0.Final-redhat-4.jar:6.1.0.Final-redhat-4]
    at org.infinispan.remoting.transport.jgroups.CommandAwareRpcDispatcher$4.run(CommandAwareRpcDispatcher.java:270) [infinispan-core-6.1.0.Final-redhat-4.jar:6.1.0.Final-redhat-4]
    at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145) [rt.jar:1.7.0_65]
    at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615) [rt.jar:1.7.0_65]
    at java.lang.Thread.run(Thread.java:745) [rt.jar:1.7.0_65]
{noformat}

The exception appears because the node had a previous consistent hash, but it was not a member of it (perhaps after a merge). The ISPN-4852 fix avoids the IllegalArgumentException by replacing the direct call to DefaultConsistentHash.getSegmentsForOwner() with a call to getOwnedSegments().

>  Backup segments may not be removed after failed state transfer
> ---------------------------------------------------------------
>
>                 Key: ISPN-5268
>                 URL: https://issues.jboss.org/browse/ISPN-5268
>             Project: Infinispan
>          Issue Type: Bug
>          Components: Core
>            Reporter: Tristan Tarrant
>            Assignee: Dan Berindei
>             Fix For: 7.0.0.Final
>
>
> Dist cache can lost data if the primary and all backups are lost at the same time.  To detect such data loss, they are monitoring the tatal sum of entries from all nodes using JMX.  If the sum reduces by a considerable ammount, data loss (by lost a segment) is concluded.  But they found a case that the sum increases, not decreases, after failed state transfer.  This is probably caused by an extra backup segment that hasn't been cleaned up when the last state transfer failed.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)