Skip to content

Latest commit

 

History

History
178 lines (121 loc) · 5.96 KB

Cluster.md

File metadata and controls

178 lines (121 loc) · 5.96 KB

Running on a cluster of nodes

EventStore supports running on multiple nodes as either a distributed Erlang cluster or as multiple single instance nodes.

Event publication

PostgreSQL's LISTEN / NOTIFY is used to pub/sub event notifications.

A single listener process will connect to the database to listen for events when using a distributed cluster. Events will be broadcast from the single listener process to a GenServer process running on each connected node that forwards events to its local subscribers. This limits the number of database connections to at most the number of running clusters.

Running EventStore on multiple nodes that are not connected together to form a cluster will result in one listener process and database connection per node.

Subscriptions

PostgreSQL's advisory locks are used to limit each uniquely named subscription to run at most once. This prevents multiple instances of a subscription from running on different nodes. Advisory locks are faster than table locks, are stored in memory to avoid table bloat, and are automatically cleaned up by the server at the end of the session.

Automatic cluster formation

You can use libcluster to automatically form clusters of Erlang nodes, with either static or dynamic node membership.

You will need to include libcluster as an additional dependency in mix.exs:

defp deps do
  [{:libcluster, "~> 2.2"}]
end

Then configure your preferred cluster topology in the environment config (e.g. config/config.exs). An example is shown below using the standard Erlang epmd daemon strategy:

config :libcluster,
  topologies: [
    example: [
      strategy: Cluster.Strategy.Epmd,
      config: [hosts: [:"[email protected]", :"[email protected]", :"[email protected]"]],
    ]
  ]

Please refer to the libcluster documentation for more detail.

Starting a cluster

  1. Run an Erlang Port Mapper Daemon (epmd):

    $ epmd -d
  2. Start an iex console per node:

    $ MIX_ENV=distributed iex --name [email protected] -S mix
    $ MIX_ENV=distributed iex --name [email protected] -S mix
    $ MIX_ENV=distributed iex --name [email protected] -S mix

The cluster will be automatically formed as soon as the nodes start.

Static cluster topology and formation

Instead of using libcluster you can configure the :kernel application to wait for cluster formation before starting your application during node start up. This approach is useful when you have a static cluster topology that can be defined in config.

The sync_nodes_optional configuration specifies which nodes to attempt to connect to within the sync_nodes_timeout window, defined in milliseconds, before continuing with startup. There is also a sync_nodes_mandatory setting which can be used to enforce all nodes are connected within the timeout window or else the node terminates.

Each node requires its own individual configuration, listing the other nodes in the cluster:

# node1 config
config :kernel,
  sync_nodes_optional: [:"[email protected]", :"[email protected]"],
  sync_nodes_timeout: 30_000

The sync_nodes_timeout can be configured as :infinity to wait indefinitely for all nodes to connect. All involved nodes must have the same value for sync_nodes_timeout.

The above approach will only work for Elixir releases. You will need to use Erlang's sys.config file for development purposes.

The Erlang equivalent of the :kernerl mix config, as above, is:

% node1.sys.config
[{kernel,
  [
    {sync_nodes_optional, ['[email protected]', '[email protected]']},
    {sync_nodes_timeout, 30000}
  ]}
].

Starting a cluster

  1. Run an Erlang Port Mapper Daemon (epmd):

    $ epmd -d
  2. Start an iex console per node:

    $ MIX_ENV=distributed iex --name [email protected] --erl "-config cluster/node1.sys.config" -S mix
    $ MIX_ENV=distributed iex --name [email protected] --erl "-config cluster/node2.sys.config" -S mix
    $ MIX_ENV=distributed iex --name [email protected] --erl "-config cluster/node3.sys.config" -S mix

The node specific <node>.sys.config files ensure the cluster is formed before starting your application, assuming this occurs within the 30 seconds timeout.

Once the cluster has formed, you can use your event store module from any node.

Usage

Using the event store when run on a cluster of nodes is identical to single node usage. You can subscribe to a stream, or all streams, on one node and append events to the stream on another. The subscription will be notified of the appended events.

Append events to a stream

alias EventStore.EventData
alias MyApp.EventStore

defmodule ExampleEvent do
  defstruct [:key]
end

stream_uuid = UUID.uuid4()

events = [
  %EventData{
    event_type: "Elixir.ExampleEvent",
    data: %ExampleEvent{key: "value"},
    metadata: %{user: "[email protected]"},
  }
]

:ok = EventStore.append_to_stream(stream_uuid, 0, events)

Read all events

alias MyApp.EventStore

recorded_events = EventStore.stream_all_forward() |> Enum.to_list()

Subscribe to all Streams

alias MyApp.EventStore

{:ok, subscription} = EventStore.subscribe_to_all_streams("example-subscription", self(), start_from: :origin)

receive do
  {:subscribed, ^subscription} ->
    IO.puts("Successfully subscribed to all streams")
end

receive do
  {:events, events} ->
    IO.puts("Received events: #{inspect(events)}")

    :ok = EventStore.ack(subscription, events)
end