I am prototyping with Infinispan 5.1B4. I am currently using cluster of two physical nodes. I notice that infinispan often gets unstable when one of the nodes leaves the cluster. All I am doing is putting values in the cache from several threads on both nodes.
The typical exceptions I see on node A when node B leaves are:
org.infinispan.util.concurrent.TimeoutException: Replication timeout for xxxx-53130
Unable to acquire lock after [10 seconds] on key [xxxx-40266.7958] for requestor [Thread[AAAAAAAAAA-4,5,main]]! Lock held by [(another thread)]
Node B leaves by simply terminating execution after job is done. I tried stopping the caches or the managers but it seemed to cause more erratic behaviour.
Is this normal? What is the graceful way for a node to leave the cluster?