How to programatically create a JMS Queue in a Clustered Environment

[seniorteaboy]

Currently we are working on moving our application from a standalone configuration to a domain / clustered configuration.

Our client app is Swing based, and talks to various EJBs in JBoss..

At the point a user logs-in, we currently create a new JMS Queue using the JBoss Native Interface - which works well for us...

however - this only works some times in a clustered environment -- since the JBoss Native Interface code, seems to have to be run on the master node / domain controller,

our code looks like this :

              ModelControllerClient client = null;

        final ModelNode op = new ModelNode();

        final ModelNode address = op.get(ClientConstants.OP_ADDR);

        address.add("profile", "full-ha");

        address.add("subsystem", "messaging");

        address.add("hornetq-server", "default");

        // the name of the queue

        address.add("jms-queue", queueName);

        // the JNDI entries

        final ModelNode entries = op.get("entries");

        for (final String jndiName : jndiNames) {

            entries.add(jndiName);

        }

        op.get(ClientConstants.OP).set(ClientConstants.ADD);

        final ModelNode result = execute(op);

and we can also do the same thing using jboss-cli.sh :

        /profile=full-ha/subsystem=messaging/hornetq-server=default/jms-queue=newQueue:add(entries=["jms/queue/newQueue"])

however -- if we try to execute either of these, on a node which is NOT the Domain Controller - we get the following response :

{

    "outcome" => "failed",

    "failure-description" => "JBAS010849: Operation add for address [

    (\"profile\" => \"full-ha\"),

    (\"subsystem\" => \"messaging\"),

    (\"hornetq-server\" => \"default\"),

    (\"jms-queue\" => \"newQueue5\")

] can only be handled by the master Domain Controller; this host is not the master Domain Controller",

    "rolled-back" => true

}

The JBoss Native Interface code is executed from a SLSB, which can of course be run on any node in the cluster (which is great from load balancing / HA point of view)..

How can we make this work on any node in the cluster ?

Or is there a way to pin an EJB to a specific node in the cluster ? (e.g. the Domain Controller),

but still have it continue to work if the Domain Controller should go down ? (e.g. have it move to another node in that scenario ?)

[mnovak]

What about just to connect to domain controller by the model controller client like:

client = ModelControllerClient.Factory.create(InetAddress.getByName("<domain_controller_ip>"), 9999);  

[seniorteaboy]

We could do this -- but of course - this might create a single point of failure ..

what should we do if the node which is running the Domain Controller goes down ?

(will/or is there a way to make the Domain Controller startup on another node ?)

[mnovak]

If you have configured slave for domain controller then it's not hard to try slave domain controller in case that master is down.