Uploaded image for project: 'JBoss Enterprise Application Platform'
  1. JBoss Enterprise Application Platform
  2. JBEAP-16195

(7.2.z) JBTM-3047 - Suspending recovery manager causes deadlock when acive RecoveryMonitor scan request exists

    Details

      Description

      This could be simulated for WFLY when it should be stopped. WildFly :shutdown command causes suspend being invoked and then JVM is about to be closed.

      Periodic recovery is active and running - for processing both phases it needs more than 20 seconds which is the default socket timeout for PeriodicMonitor call
      Client requests SCAN, the default timeout for the socket is 20 seconds[5]
      java -cp modules/system/layers/base/org/jboss/jts/main/narayana-jts-idlj-5.9.0.Final.jar com.arjuna.ats.arjuna.tools.RecoveryMonitor -host localhost -port 4712
      ServiceWorker is started and waits
      20 seconds elapsed, client gets "java.net.SocketTimeoutException: Read timed out"
      (recovery needs to be still in progress while stopping the server)
      server is asked to stop ./bin/jboss-cli.sh --command=':shutdown()' -c which causes suspend is called on the recovery manager and we have the deadlock

        Gliffy Diagrams

          Attachments

            Issue Links

              Activity

                People

                • Assignee:
                  spyrkob Bartosz Spyrko-Smietanko
                  Reporter:
                  spyrkob Bartosz Spyrko-Smietanko
                • Votes:
                  0 Vote for this issue
                  Watchers:
                  2 Start watching this issue

                  Dates

                  • Created:
                    Updated:
                    Resolved: