From: Greg W. <gr...@mo...> - 2003-10-05 11:27:20
|
There are many answers to "the clustering" question. The main reason for this is that there are lots of different reasons to cluster and lots of different webapps. Depending on your app you can cluster with: 0) Just mod_jk and no shared state between nodes 1) Shared state in DB and entity beans 2) Shared state in clustered session beans 3) Shared state in clustered http sessions 4) A combination of the above. Note that shared state is for availability - not scalability. But if you use commit option B and only stateless sesson beans, then you can probably get away without actual clustering - but you have to check that the inefficiencies of option B to not erode any gains you get from load balancing. cheers Rafal Kedziorski wrote: > hi, > > some weeks ago I asked about JBoss clustering. Now we starting implemnet > our new projekt. We use stateless session beans and entity beans (Commit > Option B). This are beans which aren't synchronized in an cluster. At > the front we have Apache Webserver with mod_jl load balancer. All > requests to the webcotainer will be redirect to two Jetty instances, > which runs with JBoss in one instance. That means, that we have two > JBoss instances. Each on own hardware. The Webcontainer uses normal JNDI > for calling sesssion beans. We don#t need HA-JNDI and we don't need > cluster. If we use session replication in Jetty, than this should work > without cluster. Or I'm wrong? > > What happens if we use MDB or Scheduler? need we than a cluster? > > > Regards, > Rafal > > > ------------------------------------------------------- > This sf.net email is sponsored by:ThinkGeek > Welcome to geek heaven. > http://thinkgeek.com/sf > _______________________________________________ > jetty-discuss mailing list > jet...@li... > https://lists.sourceforge.net/lists/listinfo/jetty-discuss > |