Hazelcast is a dropin replacement for the Play! Framework cache. Hazelcast provide also some other services like Clustered Executors, Maps, AtomicNumbers, Topics, Queues, etc...
2012-04-19 11:12:27,784 WARN pool-2-thread-49 /10.238.63.179:47600 [wonderbook] Removing connection to endpoint Address[10.238.63.179:47602] Cause => java.net.ConnectException {Connection refused}, Error-Count: 49
appearing in my log.
In my setup Hazelcast is not using multicast. I am passing Hazelcast a list of IP addresses of other machines in a cluster. The base port on all machines is 47600. The machine where the warning is appearing is a one machine cluster, it is a test machine, there will never be a second or third machine in its cluster.
It appears to be caused by SplitBrainHandler checking once a minute for other machines in a cluster that have been disconnected due to a network fault.
It appears SplitBrainHandler is attempting to periodically create a socket to ports 47601 and 47602 and failing, causing the warning. I assume it is doing this to look for other machines in the cluster that may have temporarily been disconnected from the network.
I don't believe this warning message should be logged as a warning. SplitBrainHandler is correctly doing its job looking for other machines to connect to, the fact that it cannot is a valid situation if there is no "split brain" or disconnected machines for it to find.
I keep seeing the warning message
2012-04-19 11:12:27,784 WARN pool-2-thread-49 /10.238.63.179:47600 [wonderbook] Removing connection to endpoint Address[10.238.63.179:47602] Cause => java.net.ConnectException {Connection refused}, Error-Count: 49
appearing in my log.
In my setup Hazelcast is not using multicast. I am passing Hazelcast a list of IP addresses of other machines in a cluster. The base port on all machines is 47600. The machine where the warning is appearing is a one machine cluster, it is a test machine, there will never be a second or third machine in its cluster.
It appears to be caused by SplitBrainHandler checking once a minute for other machines in a cluster that have been disconnected due to a network fault.
It appears SplitBrainHandler is attempting to periodically create a socket to ports 47601 and 47602 and failing, causing the warning. I assume it is doing this to look for other machines in the cluster that may have temporarily been disconnected from the network.
I don't believe this warning message should be logged as a warning. SplitBrainHandler is correctly doing its job looking for other machines to connect to, the fact that it cannot is a valid situation if there is no "split brain" or disconnected machines for it to find.