HAConnectionFactory UseCase question

[clebert.suconic]

I have to define what's the UseCase on HAConnectionFactory.

I could either define the load balancing policy on jndi.lookup or connectionFactory.createConnection.

Let me explaing through examples:

On these examples I have serverA, B and C. Immagine subsequent calls to either jndi.lookup or factory.createConnection doing a load balancing. (Round Robbin probably)

Example I:

connectionFactory.createConnection(); << returns a connection pointing to server A
connectionFactory.createConnection(); << server B
connectionFactory.createConnection(); << server C

Example II:

ConnectionFactory factoryA = jndi.lookup(); // this connection will always point to A
ConnectionFactory factoryB = jndi.lookup(); // this connection will always point to B
ConnectionFactory factoryC = jndi.lookup(); // this connection will always point to c

We could implement it either way. At this point I have assumed Example I, but we need to define if this is the correct behavior.

[clebert.suconic]

It would be great if anyone outside of JBoss Messaging team raises his preference on this thread. what's the best option?

[starksm64]

I can encapsulate II, so its the more general and preferred approach.

[brian.stansberry]

Option II implies that the client has connected to a JNDI service that provides a proxy with load balancing behavior (e.g. HA-JNDI). That may not be the case. It is not at all a requirement to use HA-JNDI to access clustered services (e.g. clustered EJBs) -- the HA aspect of the JNDI service and the HA aspect of the service being looked up are completely orthogonal.

[timfox]

I is preferable since we get a smoother distribution of connections across machines (assuming the same smooth load balancing policy is used in both I and II)