This week's book giveaway is in the Java in General forum. We're giving away four copies of Think Java: How to Think Like a Computer Scientist and have Allen B. Downey & Chris Mayfield on-line! See this thread for details.
I have 2 Oracle 10g web cache instances setup on 2 servers (call them OWCa and OWCb) which round robin load balance to both tomcat instances (7.0.12) installed on servers running solaris 10 (call them Ta and Tb). Ta and Tb are clustered, and everything works fine: When one of the Tomcat's is shut down, the other one picks up all the traffic via the OWC's detection of the downed tomcat instance and route them to the other tomcat instance. This other tomcat instance has the session information and services the request(s). Life is good.
Question: "Application Failover" vs "Server Failover"
If I have two apps (these could be web apps, java apps, or just a plain jsp) installed on the cluster (eg. .../webapps/proj1 and .../webapps/proj2), and only one of these apps (say proj1) "goes down" or is having issues (perhaps it sends a 404 or 500 when it can't find the .jsp page), how do I get the requests to be routed to the other tomcat instance for just proj1's requests? I guess you'd call this 'application' failover rather than 'server' failover. Is there a way to do "application failover" in Tomcat?
(Normally, when I hit my jsp page in the clustered configuration, I get it to cycle between each instance with good pages returned by each. When I 'stop' one of the instances deployments (ie. running = false), thereby 'simulating' an 'app going down' on only one instance, and then refresh a bunch of times, the pages returned cycle between a 404 page and the good page.) Part of this issue involves defining exactly what "app going down" means?
How is a 'failure' detected / triggered? for a server? for an app?
Are there any good URLs to read up on these situations? Especially w.r.t. tomcat?