as with all code it's certainly possible, but you're going to face a lot of difficulties especially the JGroups versions required by Infinispan are far more advanced than what was in 4.2.x; also Infinispan requires Java 6.
Why, and what is your need? It might be possible to use it in a single application, isolating the dependencies in your EAR/SAR, but it is likely very hard to get the JBoss clustering services to work on it fully replacing JBoss Cache.
We were using jboss-cache 1.4, our application is a swing-based client we need to maintain session at the server side to evict the user if he is idle for some defined time, 3 application server we have load-balanced (no-sticky) so clustered the jboss-cache with UDP settings everything worked fine for 4 months in production suddently for the last two weeks we are facing issues like node-1 and node-2 replication stops, buddy-backup entry getting removed so session cant be maintained.
We tried moving with TCP approach with defined clusters and port, but we are facing whole lot of problem with TCP.
1. Not picking-up the data from node-2 when node-1 is restarted.
2. All cache entries created at node-2 gets removed from other nodes when node-2 restarted.
3. BuddyNotinit exception during start-up.
So we need a desperate solution for cache with reliable clustering.
1 of 1 people found this helpful
jboss-cache 1.4 might be old and not have all the new features of Infinispan, but is stable and considered reliable. I'd suggest you to try identifying the problem in your existing deployment, it might be easier than to migrate it to Infinispan.
Another strategy if you have additional reasons to prefer Infinispan over JBoss Cache would be to migrate from JBoss 4 to JBoss 7, as seven is properly integrated with Infinispan.