Assume you have 3 servers in a cluster. On your first invocation the invocation fails and the completion status = COMPLETED_NO but the failure was non definative.
This results in no change to the cluster elements and a reset of the cluster view. The reset likely causes the second invocation to the same server that handled the first unless you are using a RandomRobin load balancer.
First is the above correct?
If so wouldn't it be better to leave the cluster view untouched so that failover moves to a different server?
Lastly, where are the GenericClusteringExceptions generated from.
Any help is appreciated.
FYI, the item that is generating this request is that in my environment I am getting "Service unavailable" exceptions with n servers in a cluster, but only one of them is busy.