Details
-
Improvement
-
Status: Resolved
-
Major
-
Resolution: Duplicate
-
None
-
None
-
All
Description
This enhancement request comes from working with a network of brokers, where each node in itself has a master/slave fault tolerant configuration.
In a 2x2 setup currently you would have a failover url on the client side consisting of 4 single urls (2 masters, 2 slaves). Once a broker dies the JMS client would reconnect
using one of the 4 urls either randomly chosen or in sequence of the url definition.
That might cause the client to reconnect to a broker that belongs to the other master/slave pair. As master/slave is designed to preserve the session state etc, it would be preferable
that the client tries to connect to the new master in its own master/slave group and only if that fails, connect to another broker.
I would imagine a nested scheme of configurations, something like:
failover:(failover:(master1,slave1)?randomize=false,failover:(master2,slave2)?randomize=false)?randomize=true
instead of
failover:(master1,slave1,master2,slave2)