SetNet.Cluster 1.2.0

<p align="center"> <img src="https://raw.githubusercontent.com/Povstalez/SetNet/master/assets/icon.png" alt="SetNet" width="96"> </p>

SetNet.Cluster

Server-to-server broadcast bus for SetNet — scale out across nodes.

When one game/app server isn't enough, run several and connect them into a mesh. Each ClusterNode runs its own dedicated listener and dials the other nodes, forming a full-mesh bus that carries only cluster traffic — separate from your player-facing server. Use it to fan cross-node events out to every node: a global chat line, a "player X came online" presence update, a cache invalidation, a "shut down for maintenance" signal. Publish with Publish<T>, react with Received / On<T>.

This is a lightweight broadcast bus, not a consensus or state-replication system — think Redis pub/sub between your nodes, not Raft.

Install

dotnet add package SetNet
dotnet add package SetNet.Cluster

Usage

Give every node a unique id, a listen port, and the addresses of the other nodes as seeds. The mesh is symmetric — it doesn't matter who dials whom.

// Node "eu-1"
var node = new ClusterNode(new ClusterNodeOptions
{
    NodeId = "eu-1",
    ListenPort = 7000,
    Seeds = new[] { ("10.0.0.2", 7000), ("10.0.0.3", 7000) },  // the other nodes
});

node.Received += (fromNode, topic, body) =>
    Console.WriteLine($"{fromNode} → {topic} ({body.Length} bytes)");

// Strongly-typed subscription (body deserialized via SetNetSerializer):
node.On<ChatLine>("global-chat", (fromNode, line) =>
    BroadcastToLocalPlayers(line));

await node.StartAsync();

// Fan out to every other node:
await node.Publish("global-chat", new ChatLine { User = "alice", Text = "hi all" });

Typically each of your player servers also hosts a ClusterNode on a separate port; when a message arrives from another node you rebroadcast it to the players connected locally.

API

new ClusterNode(ClusterNodeOptions)

ClusterNode

Notes

• Reserved wire type 65501. Cluster traffic is isolated on its own listener/port, so it never collides with your application handlers.
• Full mesh. Every node connects to every other node; messages are delivered once per link. For N nodes that's N·(N-1)/2 links — fine for a handful of nodes, not for hundreds. For large fan-out put a real message broker (NATS/Redis) behind a thin adapter instead.
• Best-effort broadcast. Delivery is reliable per live link, but a node that's down when you publish simply misses the message (it reconnects afterward). There's no store-and-forward, ordering across nodes, or dedup — add a message id/version in your payload if you need idempotency.
• Secure the mesh. Cluster links are plain SetNet connections. Run them on a private network/VPC, or enable TLS on the node configs, and don't expose the cluster port publicly.
• Symmetric seeds are fine. Listing a node that also lists you just means both try to dial; duplicate links are harmless (each carries the same broadcast).

Documentation & source

• 🐙 https://github.com/Povstalez/SetNet — full module catalog in docs/MODULES.md

License

MIT