If you want to achieve fault tolerance with failover, then master/slave is the way to go. Using a network of brokers your clients can still failover to another broker instance in the cluster but the state of each broker will not get replicated across your network. Every broker will have its own state and clients that failover to a different broker instance in the network might not receive all msgs.
Broker networks are rather used for load balancing and scaling and are not/less suitable for fault tolerance and high availability. Torsten Mielke [email protected] [email protected] On Sep 7, 2011, at 10:47 PM, dcheckoway wrote: > Thanks Gary. > > I'm considering switching a master/slave setup (in which the master is > stock, out of the box, config) over to network of brokers. Based on your > reply, it sounds like I can simply reconfigure the slave -- changing it from > slaving to using a duplex networkConnector -- and I won't have to touch the > existing master. It will automatically change its role to become a member > of the 2-node network. > > My goal, fwiw, is to better utilize the 2nd node (currently just sitting > there slaving with no producers/consumers connected) and to move away from > master/slave. > > Does all of this sound sane, or am I slightly off the mark in terms of how > I'm going about it? > > Thanks!
