hotrod client topology not updated in 5.1.0 FR

[henners]

some questions

1. I have two infinispan instances in a cluster.. I kill one of them.. The client keeps reporting java.netConnection exceptions and never considers the topology changed... Is that expected behaviour ?

2. when the killed instance restarts the client gets informed that there is a new topology but that the restarted one is removed .. So then it never gets informed that both the members are in the cluster.

   This is obviously not intended behaviour.. So my question is is this known/fixed in 5.1.2 ?

Thanks

[galder.zamarreno]

Re 1. Can you confirm that the instances did cluster? What does your client code look like? i.e. what parameters do you pass to RemoteCacheManager...etc.

Re 2. Dunno, your issue does not ring a bell right now. What's your server configuration?

[henners]

Hi, thanks for your response..

I just pass the list of 2 servers to the RemoteCacheManager ip:port, ip:port 

It seems intermittent , but 9 times out of 10 restarting an instance when it restarts the cluster reforms and state is replicated  but the client looses its knowledge of the the restarted instance for good. I dont see any API to refresh the topology , presumably because its meant to be automatic... but imho it is not working.

I have a simple server configuration with asynch replication.

So it does not sound that upgrading to 5.1.2 will make any difference .. or ?

[galder.zamarreno]

Actually, we fixed a very important issue in 5.1.2 with regards of cluster topology for Hot Rod servers: https://issues.jboss.org/browse/ISPN-1860

So, can you please upgrade to 5.1.2 both server and clients and try again?