Help me understand why `:net_kernel.start` fails

I’m experimenting with libcluster and starting by setting up a simple “first node” for said cluster.

Before I even get into libcluster, though, I’m having trouble starting distributed Erlang. This is the entirety of my so-far-very-simple application:

defmodule MyApp do
  @moduledoc false
  use Application

  def start(_type, _args) do
    IO.inspect(:erlang.node(), label: "node")

    IO.puts "starting self"
    IO.inspect(:net_kernel.start([:"my_hub@localhost"]), label: "nk start")

    topologies = [
      my_swarm: [
        strategy: Cluster.Strategy.Epmd,
        config: [hosts: [:"my_hub@localhost"]]

    children = [
      {Cluster.Supervisor, [topologies, [name: MyApp.ClusterSupervisor]]}

    Supervisor.start_link(children, strategy: :one_for_one, name: MyApp.Supervisor)

And it fails in the following way:

$ mix run
Compiling 1 file (.ex)
node: :nonode@nohost
starting self

11:19:33.205 [info]  Protocol 'inet_tcp': register/listen error: econnrefused

nk start: {:error,
 {{:shutdown, {:failed_to_start_child, :net_kernel, {:EXIT, :nodistribution}}},
  {:child, :undefined, :net_sup_dynamic,
   {:erl_distribution, :start_link, [[:my_hub@localhost], false]}, :permanent,
   1000, :supervisor, [:erl_distribution]}}}

My Googling thus far has not yielded any results. What should I be looking for?

Interestingly, it appears that using a hostname other than localhost “resolves” the problem.

1 Like

Is localhost resolvable to on your system? Long-name nodes require the host to be resolvable and reachable …

1 Like