The basic clustering-related benefit of using apache/mod_jk comes when you have more than one JBoss instance. Then you need a load balancer so the client only needs to know about a single hostname. The apache server is that load balancer. You can also buy a hardware load balancer.
Thank you for your reply!
But, let me know how the web-tier clustering be done by Jboss. I have created multiple instances of Geronimo on a single machiene, so that different Geronimo instances run with different port numbers.
But I dint find some thing similar type in Jboss as in the documentation given. Please let me know where can I find the steps of doing it.
We are running currently a product related to Information Security in Tomcat. As we need to handle huge number of transactions, we want to make use of clustering support. So, we are evaluating different application servers like Jboss and Geronimo.
Please guide me, if the web - tier clustering will be the suitable choice for my requirement or any other better approch is available with Jboss. If so, let me know that.
I have gone through that document, http://wiki.jboss.org/wiki/ConfiguringMultipleJBossInstancesOnOneMachine
The document is suggesting that when we "provide alternate ports for each instance of JBoss AS" for making the web-clustering is not recomended for a production server.
So, if I choose the other alternative, May I know what is meant by "Multihoming the network card?"
If you need to handle a very large number of concurrent web requests, yes, you'll need to use web tier clustering. Whether you also need to use some kind of state replication depends on whether you want to support HA -- i.e. user doesn't lose session state if their session moves from one server to another.
What kind of state replication depends on where your application stores any session state. Most common is in the HttpSession, so you enable web session state replication by adding the "distributable" tag to your web.xml.
Multihoming the network card means assigning more than one IP address to the same network interface. See your OS docs for how to do this. An alternative is to have a second network interface. Main point is you have a distinct IP address available for each JBoss instance on the machine.
Yes, HA is also an essential feature for us and so we approched clustering. Please finalise the option which suits best for my requirement. Basically clustering support will be seen benificiery once the product is under production. So, up to me the second option(alternate ports for each instance of JBoss AS) is not suitable option according to the document, http://wiki.jboss.org/wiki/ConfiguringMultipleJBossInstancesOnOneMachine
I would be very happy, if you find some time to reply me at the earliest.
Hi Brian Stansberry ,
Basically, I want to check the clustering support between different physical machines. I want to make a similar thing like a load balancer be on one machine and the application running on jboss on other machines.
I have followed document from jboss4-clustering, 1.5. HTTP Services. I have used apache server(httpd 2.0.63) as load balancer. But, I have not tested on multiple physical machines yet. I want to confirm with you that, will the single line in workers.properties file be sufficient to register both jboss instances to be load balanced?
Will 2 systems be sufficient to test the clustering feature?