WARNING: This is an example supplied by a user. Creating a durable subscription without specifying the client id is bad. TODO: Fix this example
How to configure MDBs to subscribe to a topic in a cluster?
Asynchronous messaging is useful for invalidation purpose across a cluster.
Let's imagine a possible scenario composed of 2 nodes running in a cluster.
hajms runs the JMS server in HA-Singleton mode. This server hosts the topic.
node1 runs an MDB with durable subscription on the topic
the JMS client publishes messages to the topic
node1 hajms MDB <--------- topic ^ | | JMS client
Each node receives a copy of the message. This "notification" may then be use to trigger cached data invalidation or a configuration reload.
Setup the environment
Installation
Create a home directory. For now on we'll call it {wiki.dir}
Download and unpack JBossAS 4.0.1SP1 in {wiki.dir}
Get the attached package hajms-clusteredtopic.zip and explode it into {wiki.dir}
/{wiki.dir}
/hajms-clusteredtopic
/jboss-4.0.1sp1
Create three copies of the all server configuration as follows
/jboss-4.0.1sp1 /server/hajms /server/node1
To make this demo more fun, we will bind each instance to a separate ip address. So first step is to configure those ips.
Here is what I've got on my box
hajms --> 192.168.4.150 node1 --> 192.168.4.151
Configuration
N/A
Deploying the mdb
On node1, the mdb makes a durable subscription anonymously to the topic bound in HA-JNDI as topic/testDurableTopic.
ejb-jar.xml ... <enterprise-beans> <message-driven> <ejb-name>DurableTopicBean</ejb-name> <ejb-class>jmslab.mdb.DurableTopicBean</ejb-class> <transaction-type>Container</transaction-type> <message-driven-destination> <destination-type>javax.jms.Topic</destination-type> </message-driven-destination> <subscription-durability>Durable</subscription-durability> </message-driven> </enterprise-beans> ...
jboss.xml ... <enterprise-beans> <message-driven> <ejb-name>DurableTopicBean</ejb-name> <destination-jndi-name>topic/testDurableTopic</destination-jndi-name> <configuration-name>Standard Message Driven Bean</configuration-name> </message-driven> </enterprise-beans> ...
Let's deploy the mdb on both nodes
cd to /hajms-clusteredtopic/src/build and type
$ant deploy
A deployment called MDB.jar is created and copied over to each node (/deploy directory)
Starting the cluster
Open 2 separate command line windows and start both node1 and hajms
$sh run.sh -c hajms -b 192.168.4.150 $sh run.sh -c node1 -b 192.168.4.151
Running the client
It's time to publish some messages to the topic
Open another command line window and cd to /hajms-clusteredtopic/src/build
$ant run-failover
On each instance, you should see the messages behing processed
16:53:50,312 INFO [DurableTopicBean] received message ID:7-11169464302961, payload: 0 16:53:51,312 INFO [DurableTopicBean] received message ID:7-11169464313122, payload: 1 16:53:52,312 INFO [DurableTopicBean] received message ID:7-11169464323123, payload: 2 16:53:53,312 INFO [DurableTopicBean] received message ID:7-11169464333124, payload: 3 16:53:54,312 INFO [DurableTopicBean] received message ID:7-11169464343125, payload: 4
Explanations
coming soon...
NOTE:
This procedure can not be used as-is with JBossAS 4.0.2. Because of a bug in HA-JNDI auto-discovery, the client is not able to look up the topic.
Until this issue is fixed, you should use the following JNDI properties
java.naming.factory.initial=org.jnp.interfaces.NamingContextFactory #java.naming.provider.url=192.168.4.150:1100 <--- comment out this line and update it accordingly java.naming.factory.url.pkgs=org.jboss.naming
Second_file:
The second attached file hajms-clusteredtopic.tar.bz2 includes a SLSB with a method postMessage(int i). This is included to demonstrate that you can post messages from within a cluster. The file also includes a SLSB client that can be run using ant
$ant run-slsbclient
The original jms client ant target has been changed from run-failover to
$ant run-mdbclient
This file also brings the MDB config up to J2EE 1.4.
Comments