Doubt about clustering fail over.

[leoliveira82]

Hello, I am trying to run the clustering service in jboss 3.2.6 and i have one problem to solve: How can i call the cluster? i mean:
I know how to call an specific machine in the cluster.
For example:

Machine 1: http://10.71.198.20:8080/project/jsp/AddressMnt.jsp?ACAO=EXIBIRFLT
Machine 2: http://10.71.198.21/project/jsp/AddressMnt.jsp?ACAO=EXIBIRFLT , and so on...

But i think that i have to call the cluster instead of calling the specific machine.
For example: http://cluster_address/projeto/jsp/EnderecoMnt.jsp?ACAO=EXIBIRFLT.
This way, i think that the fail over process is possible.
Who can help me?

[darranl]

Have you had a look at the clustering documentation available from the 'Documentation' link above.

You are going to need either a software or hardware based load balancer that your clients can connect to. That in turn will be responsible for passing the calls to the two nodes in your cluster.

[leoliveira82]

But how jboss provide the fail over mechanism?
I thought that i could get fault tolerance only with the configuration of the xmls in jboss.
Please, what can i do to obtain fail over in jboss?

[darranl]

The automatic fail over mechanism that JBoss provides is when you are accessing EJBs remotely. The reason that it is able to do this is because JBoss code is executing on the client so JBoss has control over where the requests are sent.

For web application JBoss has no control over the client code, this is the reason why an intermediate component is required that can route the calls from a single client to one of many nodes.

[leoliveira82]

darranl, now i understand how jboss works.
Thank you very much!