Remote Spawning of Actors experimental

Remote spawning is an extension of the dynamic spawn using run-time type names (see Adding Custom Actor Types experimental). The following example assumes a typed actor handle named calculator with an actor implementing this messaging interface named “calculator”.

void client(actor_system& system, const config& cfg) {
  auto node = system.middleman().connect(, cfg.port);
  if (!node) {
    cerr << "*** connect failed: "
         << system.render(node.error()) << endl;
  auto type = "calculator"; // type of the actor we wish to spawn
  auto args = make_message(); // arguments to construct the actor
  auto tout = std::chrono::seconds(30); // wait no longer than 30s
  auto worker = system.middleman().remote_spawn<calculator>(*node, type,
                                                            args, tout);
  if (!worker) {
    cerr << "*** remote spawn failed: "
         << system.render(worker.error()) << endl;
  // start using worker in main loop

We first connect to a CAF node with middleman().connect(host, port). On success, connect returns the node ID we need for remote_spawn. This requires the server to open a port with middleman().open(port) or middleman().publish(..., port). Alternatively, we can obtain the node ID from an already existing remote actor handle—returned from remote_actor for example—via hdl->node(). After connecting to the server, we can use middleman().remote_spawn<...>(...) to create actors remotely.