1 of 1 people found this helpful
You seem a bit confused about terms clustering, mod_cluster, domain and their scope. Put in simple terms, clustering handles session replication via infinispan/jgroups subsystems providing HA, mod_cluster registers backend server with the load-balancer and balances requests, and the domain controller manages and configures the servers.
What happens if the master fails?
Nothing happens. There is no master in clustering nor in mod_cluster. Regarding management, if the "master" is the domain controller, nothing really happens -- you just won't be able to manage the servers while the master is down.
If I am correct, then the only benefit of using a HA cluster or domain mode server would be to improve performance by splitting processing power and memory between slave nodes?
No, the point of using HA profiles is to get HA capabilities. The point of using domain mode is streamlined management.
Regarding domain master itself being HA, take a look at some of the Jiras:
Radoslav, thank you very much for your explanation.
Indeed I was confused regarding the difference between using domain and mod_cluster.
With your explanation I understood that domain means that we are using the master node to manage the configuration of the server groups and all slaves connected to it. On the other hand, clustering is when you use mod_cluster to balance requests to the entire system. It doesn`t have really anything to do with the management of the configuration.
Thank you very much. I have also read this topic which clarified even more the concept. Setting multiple mod_cluster load balancers in clustering environment - JBoss AS 7.1 - Project Documentation Editor