ActiveMQ
  1. ActiveMQ
  2. AMQ-1327

Unable to rejoin a failed broker to the broker network

    Details

    • Type: Bug Bug
    • Status: Closed
    • Priority: Major Major
    • Resolution: Incomplete
    • Affects Version/s: 5.2.0
    • Fix Version/s: NEEDS_REVIEW
    • Component/s: Broker
    • Labels:
      None
    • Environment:

      Cygwin in Windows XP; JDK 1.6.0_01-b06, ActiveMQ 5.0 snapshot (rev. 551727)

      Description

      Toplogy:
      A network of brokers (3 or more brokers) in a star configuration, each broker interconnected to all other brokers in the network.

      Discovery:
      This problem happens with both static connection and auto-discovery through multi-cast

      Broker Conf File: All brokers have a simple configuration as the one below:
      <beans>
      <broker brokerName="sender" persistent="false" useJmx="false" xmlns="http://activemq.org/config/1.0">
      <transportConnectors>
      <transportConnector uri="tcp://localhost:62001" />
      </transportConnectors>
      <networkConnectors>
      <networkConnector uri="static:(tcp://localhost:62002,tcp://localhost:62003,tcp://localhost:62004)"/>
      </networkConnectors>
      <persistenceAdapter>
      <memoryPersistenceAdapter/>
      </persistenceAdapter>
      </broker>
      </beans>

      How to reproduce:
      1. Start three brokers with names A, B, and C at port 62001, 62002, and 62003 respectively.
      2. Kill broker C (or any of the three brokers) via CTRL-C. Observe the other borkers now gets into recovery mode trying to reconnect to C
      3. Restart C. Notice one of both of the remaining brokers (i.e. A and B) now throws InvalidClientIDException exception. If it does not happen the first
      time, repeat until the exception occurs.

      When the exception happens, C can no longer join the broker network. The only way to recover is to shutdown and restart all three brokers.

      This does not seem to happen when there are only two brokers involved. It happens predictably in my environment when more then two brokers are involved.

        Activity

        Hide
        Helena Edelson added a comment -

        I am wondering if the reproduction of this is not equal to the original error? What I mean is, is a true broker failure testable as a broker shutdown?

        Show
        Helena Edelson added a comment - I am wondering if the reproduction of this is not equal to the original error? What I mean is, is a true broker failure testable as a broker shutdown?
        Hide
        Gary Tully added a comment -

        I recall an issue resolution that made 5.3 w.r.t the timing of bridge shutdown events. It would be interesting to know it this issue persists with 5.3.0 and if so a stack trace of the InvalidClientIDException would be most useful. thx.

        Show
        Gary Tully added a comment - I recall an issue resolution that made 5.3 w.r.t the timing of bridge shutdown events. It would be interesting to know it this issue persists with 5.3.0 and if so a stack trace of the InvalidClientIDException would be most useful. thx.
        Hide
        Timothy Bish added a comment -

        No test case provided and reported version is very old. A test case to run against a current broker would be required if this is still an issue,

        Show
        Timothy Bish added a comment - No test case provided and reported version is very old. A test case to run against a current broker would be required if this is still an issue,

          People

          • Assignee:
            Unassigned
            Reporter:
            Kevin Su
          • Votes:
            1 Vote for this issue
            Watchers:
            2 Start watching this issue

            Dates

            • Created:
              Updated:
              Resolved:

              Development