Skip to main content

​
Overview

In clustered mode, multiple nodes form a cluster. Actors are location-transparent: you send to a PID, and the framework routes to the correct node. Discovery backends (Consul, etcd, Kubernetes, NATS, mDNS, static) tell the cluster how to find peers.

​
When to use

  • High availability and horizontal scaling
  • Actor distribution across machines
  • Cluster singletons and relocation

​
Key components

  • Discovery — Pluggable provider for peer discovery
  • Cluster registry — Distributed map (Olric) for actor/grain placement
  • Remoting — TCP-based message transport between nodes

​
Configuration

Configure the actor system with WithCluster(clusterConfig) and a discovery provider. Configure remoting with WithRemote(config). The cluster joins membership, and actors can be looked up via ActorOf across nodes. See Service Discovery for provider options.

​
Relocation

When a node leaves, the leader relocates its actors and grains to remaining nodes. Singleton actors move to the leader; others are distributed. Actors can opt out of relocation via spawn options. See Relocation for the full flow, configuration, and relocatability requirements.