Therefore you could use a HASingleton. This singleton only runs on one node but is deployed on all other nodes via the deploy-hasingleton directory in order to provide failing over in case of a crash of the singleton-running server.
note that if you just implement a common singleton you should experience that this one will run on all of the cluster nodes as its deployed on all nodes via the deploy directory. could be possible that this is not a problem in your case but i wouldn't expect it to be.
i just experienced on my own ;)