1 of 1 people found this helpful
I think it is difficult to have a 'best practice' for clustering.
It depends to the application and the requirements.
You must decide between failover and performance, e.g. if you will have a full failover the performance is less because of replication.
For our application we decide to have only a transactional cache and a stickyness to one node during the transaction (it is mostly a EJB based app).
For our requirements it works.
The plan is to have about 70% load on each node to have a good GC behaviour and being able to handle peaks.
Thanks for the answer. I find myself being at a disadvantage alot with jboss because i am not the developer and the knowldge of clustering and admin I have is far beyond that of the vendor(which isn't saying much). I have read through the clustering guide more than once which I find to be missing some info that I have had to gather from other more generous tech blogs, but non of them have given me enough info to be more than just dangerous. I know there wont be one definitive guide for clustering but I was hoping for something stronger than the online clustering doc and the book published by Jboss, inc.
Any solid referance you can direct me to would be very helpful.