1 Reply Latest reply on Feb 11, 2013 5:37 AM by Andy Taylor

    live.stop() issues and proposed patch

    borges Novice

      Note that the patch is not even a PR at this moment...

       

      We used to have problems where:

      1. live.stop() is called;
      2. network went down;
      3. live would hang because of missing replication tokens.

       

      We scheduled a thread to clean replication tokens in 10s. But the trick is that right after that we would call replicationManager.stop() which would clean them all anyway.

       

      [...]

       

      besides that there were the following problems in server.stop() (related or not to replication.)

      • calling storageManager.closeIdGenerator() very early. Before closing the server sessions.
      • sending LIVE IS STOPPING before closing server sessions.
      • not passing criticalIoError to storageManager.stop(boolean criticalError), which defeats the point of having the method there.

       

       

      @Clebert, the patch above fixes ReplicatedPagedFailoverTest.testFailThenReceiveMoreMessagesAfterFailover2() for me (man, jboss.org forum interface sucks...)

       


      @Clebert and @AndyTaylor can you guys take a look at https://github.com/FranciscoBorges/hornetq/commit/901d2173be390861a2055b51bcd46cc63c91b58c for obvious problems before I send a PR? (I would rather avoid merging this without a proper discussion).