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

Answering the Core Questions

1. Transitioning to a static member list without touching Vert.x code.

• Keep your Vert.x application unchanged; only the Hazelcast configuration needs updating.
• In hazelcast.xml replace the <join><multicast> section with a <join><tcp-ip> block that lists every instance’s IP:port or DNS name.
• Example snippet:

<network>
  <join>
    <multicast enabled="false"/>
    <tcp-ip>
      <member>10.0.0.1:5701</member>
      <member>10.0.0.2:5701</member>
      <member>10.0.0.3:5701</member>
    </tcp-ip>
  </join>
</network>

Deploy this file to every JVM (or use a shared config service). Vert.x will pick it up automatically when the Event Bus starts.

2. Event Bus timeout behavior.

• Clustered mode: Messages are routed through Hazelcast’s TCP overlay. If a target node is unreachable, Hazelcast marks it as failed after the heartbeat interval (default ~30 s). The Event Bus will then route the message to another node or, if none are available, trigger the send(..., timeout) callback after the specified timeout.
• Standalone mode: There is no distributed heartbeat. A send to a non‑existent address fails immediately, and the timeout callback fires almost instantly (within a few milliseconds).

Thus, in a cluster you may see a longer delay before a timeout fires, whereas in standalone you see an almost immediate failure.

Recommended Steps for Cloud Deployment

1. Verify that multicast (UDP 5431–5433) is blocked in your VPC or security group. Use nc -zv -u 224.0.0.1 5431 from a VM.
2. Create a hazelcast.xml with TCP‑IP join as shown above. If you use dynamic IPs (e.g., EC2 auto‑scaling), expose the private DNS names instead.
3. Place the file in the same location as your Vert.x launcher (or point -Dhazelcast.config=…).
4. Start all instances; check Hazelcast’s web console or JMX to confirm the member list matches.
5. Run a simple eventBus.send("test.address", "msg", 5000, handler) test to observe the timeout behavior. Log the callback timestamp.
6. If you need tighter control over heartbeat or timeout, adjust hazelcast.network.heartbeat.interval and vertx.eventBus.timeout in the same config.

Diagnostic Detail Needed

To fine‑tune the timeout alignment, could you confirm which Vert.x and Hazelcast major versions you are using? The default heartbeat interval and timeout values differ between 4.x and 5.x releases.