Message redundancy with Cluster Node failure

I am trying to setup a 2 node cluster in order to have load balancing and failure protection. I have the cluster setup but in testing I am trying to test failure protection by shuttingdown one node then bring it back up. When I do this, the messages sent to the other Node (the one that was alive the whole time) will not appear on clients connecting to the node I just brought back online. How can I get messages to resync from to node that was alive the whole time?

explaination break down:

2 nodes A and B,

2 websites that connect to them AA, BB,

AA->A, BB->B,

bring down B,

refresh page and connect BB->A,

sends to messages,

restart B,

refresh page so BB->B,

BB client does not see messages sent while B was down

Current setup:
2 Openfire 4.1.1 on two different Windows 2012 servers clustered with Hazelcast plugin

connect via Openfire’s internal BOSH connection manager

using strophe.js to build custom front-end in asp.net