Failover logic in ClusteredConnectionFactoryDelegate

[timfox]

I notice that quite a bit of the failover logic has ended up in the ClusteredConnectionFactoryDelegate class.

The idea with the client aop stack is all the logic - including the failover logic should be in a set of aspects and the actual delegate classes should be as dumb and thin as possible.

So this logic should really be in a FailoverAspect, or something similar.

[clebert.suconic]

ClusteredConnectionFactoryDelegate is the class responsible for LoadBallancing ConnectionFactories.

The only failover logic on it is the handleConnectionException which is just recreating the connection and calling ClientConnectionDelegate::failOver.

When we start to listen exceptions on other interceptors I might refactor this.

[timfox]

There is quite a lot of logic in there.

There is the load balancing policy - which is currently hardcoded to be round robin - needs to be separated out and made pluggable.

As I say, this should be in an aspect. If we don't do it now, it will probably never get done and the code will drift into an unmaintainable mess.