failover

[monasehgal]

Hi
I'm doing server clustering and i've 2 jboss-3.0.6 _tomcat-4.1.18 servers running on 2 different windows m/c.
When i start the two servers both of them start without giving any errors and (which was started later specifies having one more member in the cluster).
Now i start my application on both the servers giving the url as
//localhost:8080/myapplication

both the application start.
Now if i shut down the server on m/c 1 then the application doesnot work (it should be able to perform failover and should be able to locate the other server)and displays
Page not found error
And also the the server on m/c 2 displays 1 dead member .
right now my jndi.properties file is -
java.naming.factory.initial=org.jnp.interfaces.NamingContextFactory
java.naming.factory.url.pkgs=org.jboss.naming:org.jnp.interfaces
java.naming.provider.url=localhost:1100

I'm also attaching my cluster-service.xml file
Any suggestions and help would be appriciated.

Thanx

[heartbit]

monasehgal, are you sure that you have any kind of load balancing mechanism in front of those two JBoss/Tomcat machine? (ie. Hardware L4/L7 switch or Web Server plugin kind thing)

There is no way to get fail-over with no load balancer.

[ltam]

I'm new to JBoss and haven't experimented much with it yet, but I read in the Clustering manual that it has smart proxies that use RMI to do failover. Am I misunderstanding the documentation?

However, it does say that non-RMI clients seem to loose the clustering feature. Could this be the problem?
Refer to P.20 of the manual.

[heartbit]

Yes. that EJB stub dose the job doen *for EJB requset*.

There should be other *for HTTP request*. Something like H/W L4 switch.

[mrtom]

Does anything need to be done to use the smart proxies? We are experiencing the same problems as monasehgal.

Our problem is that there appears to be no load balancing and no failover. Whichever JBoss node starts up last seems to take all of the work (we've tried using round/random robin and first available). When the server doing all the work is shut down (cleanly) no failover happens and the client gets very unhappy. The only time I've managed to get any kind of failover is when I touch my files in the farm dir (as the second node can take a while to redeploy the new files). In this case I see the busy node undeploy it's files and start using the other node (which has already deployed the new archives).

If I do another test by pulling the network cable out of the busy node everything hangs until it becomes available again. The other server detects the loss of the node but the clients don't (this seems most worrying when I am using first available).

Just for reference we are trying this using the auto-detection so java.naming.provider.url=""

[slaboure]

to help us:
- provide your full jboss version (we expect you to use a *recent* version)
- modify one of your methods you call to simply have this in its body:
throw new Exception ();

and send the exceptions you see BOTH on the server AND on the client.

Cheers,


sacha