I have a question about the the strategy for choosing alive cluster nodes. Lets say I have a replicated Infinisapn cluster with two nodes [A,B] and a hotrod client. I am using 5.2.1 release. I do the following steps:
When I start the client it says that both nodes [A,B] are in cluster which is fine.
I kill node B
From the client I put some value in the clustered cache and on that operation the topology is updated saying that only node [A] is alive. This is fine too.
Now I fire up node B and after that kill node A.
When I try again to put some value a TransportExcetpion is thrown with no connection. (Here I want to try both node A, and node B despite that B is not in the alive nodes list)
When I fire up node A everything is fine.
So is there some way without implementing my own TCPTransportFactory to manually update topology so that the hotrod client can retry any previously dead hosts or to be notified that node B is up and node A is down so to use node B instead of node A?