3 Replies Latest reply on Jul 25, 2012 4:24 AM by ataylor

    Simple clustering question with JBoss AS7

    t3rm1

      Hello,

       

      I have a quick question which I couldn't find an answer for. I hope you don't mind helping me on this maybe easy problem

       

      I'm using JBoss 7.1 in clustered standalone mode (2 server, basicially unaltered standalone-ha.xml configuration). Here is what I want to achieve:
      Server A receives a message. I made it taking 20 seconds to finish the work on it. Before Server A can complete the message I kill the process. I now want Server B to notice the server crash and process the message.
      I tested this and it doesn't work right now (though the server crash is recognized). The message isn't being processed by Server B. If I start Server A again it will finish it's work. However this is not what I want.

       

      As far as I understand all nodes are aware of the current message queue. I know there is a HA setting where you can setup a server as a backup but I don't want this because it requires a shared store for the journal files.

       

      Is it possible to do this with a simple cluster?

        • 1. Re: Simple clustering question with JBoss AS7
          gaohoward

          I think what you need is a backup with replication-- which means no shared store required.

          However I don't think it's in AS7 now, but I believe it will soon.

           

          http://hornetq.blogspot.com/2012/07/hornetq-230-alpha-released.html

           

           

          Howard

          • 2. Re: Simple clustering question with JBoss AS7
            t3rm1

            Well, the documentation states that every node has the current state of the queue. If Server A processes a message but fails because it crashed then this message should be still in the queue, right? If this is the case then there shouldn't be a problem for Server B to process this message.

            • 3. Re: Simple clustering question with JBoss AS7
              ataylor

              Well, the documentation states that every node has the current state of the queue. If Server A processes a message but fails because it crashed then this message should be still in the queue, right? If this is the case then there shouldn't be a problem for Server B to process this message.

              yes the message is still in the queue but since the server has crashed it will never get delivered, you need to configure a backup server (using shared storage) to take over from the crashed live. Replication will be in a later version of as7