[infinispan-issues] [JBoss JIRA] (ISPN-5746) Continuous query desyncs with PreferAvailabilityStrategy

[Rich DiCroce (JIRA)]

Rich DiCroce created ISPN-5746:
----------------------------------

             Summary: Continuous query desyncs with PreferAvailabilityStrategy
                 Key: ISPN-5746
                 URL: https://issues.jboss.org/browse/ISPN-5746
             Project: Infinispan
          Issue Type: Bug
    Affects Versions: 8.0.0.CR1
            Reporter: Rich DiCroce
         Attachments: ispn.png

Continuous queries get desynchronized when PreferAvailabilityStrategy is in use (i.e. partition-handling = false) on a replicated or distributed cache, and the cluster gets partitioned. When the partition heals, the nodes in the smaller partition wipe their caches and reacquire from the larger partition. But the continuous query listener doesn't listen for the CacheEntryInvalidatedEvent that gets fired when wiping the cache, so listeners are not notified that cache entries have left the query.

The attached screenshot demonstrates the result of this series of events on a simple test WAR deployed on WildFly 10.0.0.Beta2. (I had to frankenstein the query modules in due to WFLY-4695, but I doubt that has any impact on the validity of this issue.) Notice that there are only two entries in the cache, but the continuous query (which just counts the number of entries) believes there are four.



--
This message was sent by Atlassian JIRA
(v6.4.11#64026)