Environment:

• 3 node cluster with a HA singleton-deployment
• a client with all 3 cluster nodes in PROVIDER_URL for initial context

Issue:

When the node running the singleton-deployment is shutdown/killed, an async message is sent from the new node where the singleton-deployment is started next to the client. This lead to an update of the NodeInformation held inside DiscoveredNodeRegistry. In parallel the client continues to call EJBs on the cluster. These are failing due to NoSuchEJBException on all nodes. These NoSuchEJBException lead to removing the module information on the NodeInformation again, which might override the recently received information for the new node running the HA singleton-deployment.