Google Cloud Pub/Sub

Cloud Pub/Sub is a fully-managed real-time messaging service provided by Google.

Getting Started

In order to use Broadway with Cloud Pub/Sub you need to:

Setup a Cloud Pub/Sub project
Configure your Elixir project to use Broadway
Define your pipeline configuration
Implement Broadway callbacks
Run the Broadway pipeline
Tune the configuration (Optional)

If you are just getting familiar with Google Pub/Sub, refer to the documentation to get started. Instead of testing against a live environment, you may also consider using the emulator to simulate integrating with Cloud Pub/Sub.

If you have an existing project, topic, subscription, and credentials, you can skip step 1 and jump to Configure the project section.

Setup Cloud Pub/Sub project

In this tutorial we'll use the gcloud command-line tool to set everything up in Google Cloud. Alternatively, you can roughly follow this guide by using Cloud Console.

To install gcloud follow the documentation. If you are on macOS you may consider installing it with Homebrew:

$ brew install --cask google-cloud-sdk

Now, authenticate the CLI:

$ gcloud auth login

Then, create a new project:

$ gcloud projects create test-pubsub

A new topic:

$ gcloud pubsub topics create test-topic --project test-pubsub
Created topic [projects/test-pubsub/topics/test-topic].

Note: If you run this command immediately after creating a new Google Cloud project, you may receive an error indicating that your project's organization policy is still being provisioned. Just wait a couple minutes and try again.

And a new subscription:

$ gcloud pubsub subscriptions create test-subscription --project test-pubsub --topic test-topic
Created subscription [projects/test-pubsub/subscriptions/test-subscription].

We also need a service account, an IAM policy, as well as API credentials in order to programmatically work with the service. First, let's create the service account:

$ gcloud iam service-accounts create test-account --project test-pubsub
Created service account [test-account].

Then the policy. For simplicity we add the general role roles/editor, but make sure to examine the available roles and choose the one that best suits your use case:

$ gcloud projects add-iam-policy-binding test-pubsub \
    --member serviceAccount:test-account@test-pubsub.iam.gserviceaccount.com \
    --role roles/editor
Updated IAM policy for project [test-pubsub].
(...)

And now the credentials:

$ gcloud iam service-accounts keys create credentials.json --iam-account=test-account@test-pubsub.iam.gserviceaccount.com
created key [xxx] of type [json] as [key] for [test-account@test-pubsub.iam.gserviceaccount.com]

This command generated a credentials.json file which will be useful later. Note, the IAM account pattern is <account>@<project>.iam.gserviceaccount.com. Run gcloud iam service-accounts list --project test-pubsub to see all service accounts associated with the given project.

Finally, we need to enable Pub/Sub for our project:

$ gcloud services enable pubsub --project test-pubsub
Operation "operations/xxx" finished successfully.

Configure the project

In this guide we're going to use BroadwayCloudPubSub, which is a Broadway Cloud Pub/Sub Connector provided by Dashbit.

Starting a new project

If you plan to start a new project, just run:

$ mix new my_app --sup

The --sup flag instructs Elixir to generate an application with a supervision tree.

Setting up dependencies

Add :broadway_cloud_pub_sub to the list of dependencies in mix.exs, along with the Google Cloud authentication library of your choice (defaults to :goth):

defp deps() do
  [
    ...
    {:broadway_cloud_pub_sub, "~> 0.7"},
    {:goth, "~> 1.0"}
  ]
end

Don't forget to check for the latest version of dependencies.

Define the pipeline configuration

Broadway is a process-based behaviour and to define a Broadway pipeline, we need to define three functions: start_link/1, handle_message/3 and handle_batch/4. We will cover start_link/1 in this section and the handle_ callbacks in the next one.

Similar to other process-based behaviour, start_link/1 simply delegates to Broadway.start_link/2, which should define the producers, processors, and batchers in the Broadway pipeline. Assuming we want to consume messages from the test-subscription, the minimal configuration would be:

defmodule MyBroadway do
  use Broadway

  alias Broadway.Message

  def start_link(_opts) do
    Broadway.start_link(__MODULE__,
      name: __MODULE__,
      producer: [
        module:
          {BroadwayCloudPubSub.Producer,
           subscription: "projects/test-pubsub/subscriptions/test-subscription"}
      ],
      processors: [
        default: []
      ],
      batchers: [
        default: [
          batch_size: 10,
          batch_timeout: 2_000
        ]
      ]
    )
  end

  ...callbacks...
end

For a full list of options for BroadwayCloudPubSub.Producer, please see the documentation.

For general information about setting up Broadway, see Broadway module docs as well as Broadway.start_link/2.

Note: Even though batching is optional since Broadway v0.2, we recommend all Cloud Pub/Sub pipelines to have at least a default batcher, as that allows you to control the exact batch size and frequency that messages are acknowledged to Cloud Pub/Sub, which often leads to pipelines that are more cost and time efficient.

Implement Broadway callbacks

In order to process incoming messages, we need to implement the required callbacks. For the sake of simplicity, we're considering that all messages received from the queue are strings and our processor calls String.upcase/1 on them:

defmodule MyBroadway do
  use Broadway

  alias Broadway.Message

  ...start_link...

  def handle_message(_, %Message{data: data} = message, _) do
    message
    |> Message.update_data(fn data -> String.upcase(data) end)
  end

  def handle_batch(_, messages, _, _) do
    list = messages |> Enum.map(fn e -> e.data end)
    IO.inspect(list, label: "Got batch of finished jobs from processors, sending ACKs to Pub/Sub as a batch.")
    messages
  end
end

We are not doing anything fancy here, but it should be enough for our purpose. First we update the message's data individually inside handle_message/3 and then we print each batch inside handle_batch/4.

For more information, see Broadway.handle_message/3 and Broadway.handle_batch/4.

Run the Broadway pipeline

To run your Broadway pipeline, you need to add it as a child in a supervision tree. Most applications have a supervision tree defined at lib/my_app/application.ex. You can add Broadway as a child to a supervisor as follows:

children = [
  {MyBroadway, []}
]

Supervisor.start_link(children, strategy: :one_for_one)

The final step is to configure credentials. You can set the following environment variable:

export GOOGLE_APPLICATION_CREDENTIALS=/path/to/credentials.json

See Goth documentation for alternative ways of authenticating with the API.

Now the Broadway pipeline should be started when your application starts. Also, if your Broadway pipeline has any dependency (for example, it needs to talk to the database), make sure that it is listed after its dependencies in the supervision tree.

If you followed the previous section about setting the project with gcloud, you can now test the the pipeline. In one terminal tab start the application:

$ iex -S mix

And in another tab, send a couple of test messages to Pub/Sub:

$ gcloud pubsub topics publish  projects/test-pubsub/topics/test-topic --message "test 1"
messageIds:
- '651428033718119'

gcloud pubsub topics publish  projects/test-pubsub/topics/test-topic --message "test 2"
messageIds:
- '651427034966696'

Now, In the first tab, you should see output similar to:

Got batch of finished jobs from processors, sending ACKs to Pub/Sub as a batch.: ["TEST 1", "TEST 2"]

Tuning the configuration

Some of the configuration options available for Broadway come already with a "reasonable" default value. However those values might not suit your requirements. Depending on the number of messages you get, how much processing they need and how much IO work is going to take place, you might need completely different values to optimize the flow of your pipeline. The concurrency option available for every set of producers, processors and batchers, among with max_demand, batch_size, and batch_timeout can give you a great deal of flexibility.

The concurrency option controls the concurrency level in each layer of the pipeline. See the notes on Producer concurrency and Batcher concurrency for details.

Here's an example on how you could tune them according to your needs.

defmodule MyBroadway do
  use Broadway

  def start_link(_opts) do
    Broadway.start_link(__MODULE__,
      name: __MODULE__,
      producer: [
        ...
        concurrency: 10,
      ],
      processors: [
        default: [
          concurrency: 100,
          max_demand: 1,
        ]
      ],
      batchers: [
        default: [
          batch_size: 10,
          concurrency: 10,
        ]
      ]
    )
  end

  ...callbacks...
end

In order to get a good set of configurations for your pipeline, it's important to respect the limitations of the servers you're running, as well as the limitations of the services you're providing/consuming data to/from. Broadway comes with telemetry, so you can measure your pipeline and help ensure your changes are effective.

← Previous Page Apache Kafka

Next Page → RabbitMQ