Skip to content

learn-co/railway_ipc

Repository files navigation

RailwayIpc

The purpose of Railway is to standardize a common set of API's and procedures for dealing with inter-process communication between applications via RabbitMQ. This projects implements the API's and procedures for Elixir based projects.

Installation

If available in Hex, the package can be installed by adding railway_ipc to your list of dependencies in mix.exs:

def deps do
  [
    {:railway_ipc, "~> 0.3.0"}
  ]
end

Getting Started

Configure Railway to work with your Repo. Add the following to your config/config.exs:

config :railway_ipc,
  repo: ApplicationName.Repo

Run the mix task to generate the migrations to add the published messages and consumed messages tables to your app's DB:

mix railway_ipc.generate_migrations ./path/to/migration/directory
mix ecto.migrate

Note: Path to migration directory defaults to ./priv/repo/migrations if none is passed in.

If there are issues running the migration or deploying the migration, try manually writing the name of the migration module (not the file) to avoid using interpolation.

Documentation can be generated with ExDoc and published on HexDocs. Once published, the docs can be found at https://hexdocs.pm/railway_ipc.

Consuming the same message on multiple queues

Out of the box, Railway can handle storing the same messages multiple times if it's consumed on multiple queues. If you are upgrading Railway from 2.1 or earlier, you will need to run the following migration to make uuid and queue a combined primary key for the consumed messages table.

defmodule YOUR_APP_NAME_HERE.Repo.Migrations.UpdateRailwayMessagePKey do
  use Ecto.Migration

  import Ecto.Query, only: [from: 2]
  alias Registrar.Repo

  def up do
    alter table(:railway_ipc_consumed_messages) do
      add :new_uuid, :uuid
    end

    flush()

    from(m in "railway_ipc_consumed_messages", update: [set: [new_uuid: m.uuid]])
    |> Repo.update_all([])

    alter table(:railway_ipc_consumed_messages) do
      remove :uuid
      modify :queue, :string, primary_key: true, null: false
      modify :new_uuid, :uuid, primary_key: true, null: false
    end

    rename(table(:railway_ipc_consumed_messages), :new_uuid, to: :uuid)

    create unique_index(
      "railway_ipc_consumed_messages",
      [:uuid, :queue],
      name: :railway_ipc_consumed_messages_uniqueness_index
    )
  end

  def down do
    alter table(:railway_ipc_consumed_messages) do
      add :old_uuid, :uuid
      add :old_queue, :string
    end

    flush()

    from(m in "railway_ipc_consumed_messages", update: [set: [old_uuid: m.uuid, old_queue: m.queue]])
    |> Repo.update_all([])

    alter table(:railway_ipc_consumed_messages) do
      remove :uuid
      remove :queue
      modify :old_uuid, :uuid, primary_key: true, null: false
      modify :old_queue, :string, null: false
    end

    rename(table(:railway_ipc_consumed_messages), :old_uuid, to: :uuid)
    rename(table(:railway_ipc_consumed_messages), :old_queue, to: :queue)
  end
end

For more information on this process, check out this blogpost: https://niallburkley.com/blog/changing-primary-keys-in-ecto/

Development

Setup

Requires Erlang, Elixir, a local RabbitMQ instance, a local PostgreSQL database. Clone the repo and run the bin/setup script.

Tests / Linting

This project uses ExUnit for tests and Credo for linting code.

CI Workflow

This project uses CircleCI. All pushes to any branch will trigger a build that runs specs and lints the code.

Cutting a New Release

Steps for releasing a new version:

  • update the version in mix.exs and README.md
  • update the CHANGELOG
    • create a new release heading with the new version number and date
    • create an empty [Unreleased] section with empty headings
    • update links at the bottom to reflect new version
  • commit the changes with a message like: "Prepping for release 0.3.0"
  • tag the commit with git tag x.x.x where x.x.x is the new version number
  • push the tag to GitHub git push --tags
  • publish to hex using the command mix hex.publish (Note: Since we're using a forked version of the elixir/protobuf library, we cannot publish to Hex. Once we're able to get back on the official elixir/protobuf version we'll start publishing to Hex again. We are still creating Git tags for each release, however.)