Uploaded image for project: 'JBoss Enterprise Application Platform'
  1. JBoss Enterprise Application Platform
  2. JBEAP-29376

(8.1.z) [CLUSTERING] Cache inconsistencies in fail-over tests

XMLWordPrintable

    • Icon: Bug Bug
    • Resolution: Unresolved
    • Icon: Major Major
    • None
    • 8.1.0.Beta
    • Clustering
    • False
    • Hide

      None

      Show
      None
    • False

      We spotted an issue which is apparently related to https://issues.redhat.com/browse/JBEAP-29215; they are solved when using EAP build from this branch: https://github.com/pferraro/jboss-eap8/tree/8.1.x;

      The issue is intermittent ans shows approximately on 3 test runs every 4 test runs;

      Specifically, in this scenario we have a distributed application which stores SFSB into session data; the application is deployed on a 4 nodes cluster using the TCP JGroups stack;

      The cache inconsistency show up after the EAP instance where a session was created is killed (or shut down): as expected the session is routed to a different node but after the first request to the new node is successful, the second one shows a cache inconsistency because the serial number stored in the distributed cache isn't incremented as expected; See the comment below for the details;

      In https://issues.redhat.com/browse/JBEAP-29215, the error consisted in the session being invalidated and replaced; as consequence, the serial number stored in the distributed cache, was re-initialized from 0;

      The issue itself shows up one or two times every test run and doesn't significantly affect the client fail rate; nevertheless we did not observer the issue in EAP 8.0 and hence, we are setting this issue's priority to "blocker".

              pferraro@redhat.com Paul Ferraro
              tborgato@redhat.com Tommaso Borgato
              Votes:
              0 Vote for this issue
              Watchers:
              2 Start watching this issue

                Created:
                Updated: