Vert.x Event Bus and Hazelcast: Multicast vs Static Discovery in Cloud

Cluster Manager Interoperability

Vert.x uses an Event Bus that runs in standalone mode by default. For production, clustering via a Cluster Manager such as Hazelcast extends the bus across multiple JVMs.

A common discrepancy arises between local development, where multicast discovery works on a LAN, and cloud environments (AWS, GCP) where multicast is blocked. This prevents cluster formation despite identical code.

Switching to a static member list or cloud discovery plugin resolves formation, but the behavior of the Event Bus during handshake and the serialization overhead on message latency compared to local delivery remains a critical design uncertainty.

• What is the recommended strategy for transitioning Hazelcast discovery from multicast to a static member list without changing core Vert.x application logic?
• How does the Event Bus handle message timeouts when a remote node is unreachable in a clustered configuration versus a standalone instance?