Uploaded image for project: 'JBoss Enterprise Application Platform 4 and 5'
  1. JBoss Enterprise Application Platform 4 and 5
  2. JBPAPP-6932

HASingletons continue running after merge

    XMLWordPrintable

Details

    • Bug
    • Resolution: Done
    • Major
    • EAP_EWP 5.1.2 ER1
    • EAP_EWP 5.1.1
    • Clustering
    • None
    • Hide

      Start two clustered nodes with a deployment in deploy-hasingleton.
      Split the cluster (for instance, pull the network plug).
      Reconnect the cluster.

      Expected result: HASingleton deployment is stopped on non-master node
      Actual result: HASingleton deployment continues running on both nodes

      Show
      Start two clustered nodes with a deployment in deploy-hasingleton. Split the cluster (for instance, pull the network plug). Reconnect the cluster. Expected result: HASingleton deployment is stopped on non-master node Actual result: HASingleton deployment continues running on both nodes
    • Release Notes
    • Hide
      When two cluster nodes were merged after a split, only one of them should have continued to HASingletons. In fact both continued to run HASingletons, resulting in unreliable cluster operation. The cause was an error in the logic applied when node merged. This is now resolved and cluster merges now operate as expected.
      Show
      When two cluster nodes were merged after a split, only one of them should have continued to HASingletons. In fact both continued to run HASingletons, resulting in unreliable cluster operation. The cause was an error in the logic applied when node merged. This is now resolved and cluster merges now operate as expected.
    • Documented as Resolved Issue
    • NEW

    Description

      When two cluster members don't see each other (network split), they will both be a master and run any HASingletons.
      When they merge back together, only the new master should continue running the HASingleton.
      But the HASingleton continues to run on both nodes after the merge.

      This is because DistributedReplicantManagerImpl does not notify the HASingleton of the new cluster members on a merge.
      This likely affects all other services that depend on DistributedReplicantManagerImpl notifications also.

      Root cause is DistributedReplicantManagerImpl#addReplicant's return value is backwards.
      It should "@return true, if this replicant was newly added to the map, false otherwise", but does just the opposite.

      Attachments

        Activity

          People

            rhn-support-dereed Dennis Reed
            rhn-support-dereed Dennis Reed
            Russell Dickenson Russell Dickenson (Inactive)
            Votes:
            0 Vote for this issue
            Watchers:
            1 Start watching this issue

            Dates

              Created:
              Updated:
              Resolved: