RabbitMQ Messaging & Async Systems · レッスン

Work Queues：公平なディスパッチ

ラウンドロビン方式で複数の Consumer にタスクを分配する Work Queues を実装します。時間のかかるタスクを非同期に処理する方法を学びます。

「Work Queues：公平なディスパッチ」はCoddyKit上の無料RabbitMQ Messaging & Async Systemsレッスンです。これはレッスン2/4です。下記で完全なレッスンを無料で読むことができます。その後、ブラウザ内の組み込みコードエディタと24時間対応のAIチューターでハンズオン演習できます。これはRabbitMQ Messaging & Async Systems学習パスの一部であり、ウェブとCoddyKitアプリ全体で進捗が同期されます。 RabbitMQ Messaging & Async Systemsコースには全4レッスンが含まれています。

このレッスンの一部はまだ翻訳されておらず、英語で表示されています。

Meet Work Queues

Welcome to Work Queues! In distributed systems, you often have tasks that take time to complete, like processing an image or generating a report.

Work Queues are a pattern that helps distribute these time-consuming tasks among multiple workers (consumers) efficiently, preventing any single worker from getting overloaded.

Why Use Work Queues?

Imagine you have many jobs to do, but only one employee. If all jobs go to that single employee, they'll get overwhelmed and tasks will pile up.

Load Balancing: Work queues allow you to add more workers to share the load.
Asynchronous Processing: The producer doesn't wait for a task to finish, it just adds it to the queue.
Reliability: If one worker fails, others can pick up tasks.

How Work Queues Operate

The setup for a work queue is simple:

One producer sends messages (tasks) to a single queue.
Multiple consumers listen to this same queue.
RabbitMQ ensures that each message is delivered to only one of the waiting consumers.

This way, tasks are never duplicated and are processed in parallel.

Fair Dispatch: Round-Robin

By default, RabbitMQ distributes messages to consumers using a round-robin strategy. This means messages are sent to consumers in a rotating fashion:

Consumer 1 gets the first message.
Consumer 2 gets the second message.
Consumer 1 gets the third message, and so on.

This aims for an even distribution of tasks among all active consumers.

The Task Producer

Let's create a producer that sends 10 tasks to our queue. Each task will be a simple string like 'Processing image 1'.

Run this code once to populate the queue with tasks.

import com.rabbitmq.client.Channel;
import com.rabbitmq.client.Connection;
import com.rabbitmq.client.ConnectionFactory;

public class TaskProducer {
    private final static String QUEUE_NAME = "task_queue";

    public static void main(String[] argv) throws Exception {
        ConnectionFactory factory = new ConnectionFactory();
        factory.setHost("localhost"); // Assuming RabbitMQ is local

        try (Connection connection = factory.newConnection();
             Channel channel = connection.createChannel()) {

            channel.queueDeclare(QUEUE_NAME, false, false, false, null);

            for (int i = 0; i < 10; i++) {
                String message = "Processing image " + (i + 1);
                channel.basicPublish("", QUEUE_NAME, null, message.getBytes("UTF-8"));
                System.out.println(" [x] Sent '" + message + "'");
            }
            System.out.println(" [x] All tasks sent.");
        }
    }
}

Producer Code Breakdown

What's happening in our producer code?

We establish a connection and a channel to interact with RabbitMQ.
channel.queueDeclare(QUEUE_NAME, false, false, false, null); ensures the queue exists. The false flags keep it non-durable and non-exclusive for simplicity here.
A loop sends 10 messages to the task_queue. Each message represents a distinct task.

Our First Task Consumer

Now, let's create a consumer, which we'll call a 'worker'. This worker will listen for tasks from the task_queue.

We'll simulate a long-running task using Thread.sleep(). Run this code in one terminal window.

import com.rabbitmq.client.Channel;
import com.rabbitmq.client.Connection;
import com.rabbitmq.client.ConnectionFactory;
import com.rabbitmq.client.DeliverCallback;

public class TaskConsumer {
    private final static String QUEUE_NAME = "task_queue";

    public static void main(String[] argv) throws Exception {
        ConnectionFactory factory = new ConnectionFactory();
        factory.setHost("localhost"); // Assuming RabbitMQ is local

        Connection connection = factory.newConnection();
        Channel channel = connection.createChannel();

        channel.queueDeclare(QUEUE_NAME, false, false, false, null);
        System.out.println(" [*] Waiting for messages. To exit press CTRL+C");

        DeliverCallback deliverCallback = (consumerTag, delivery) -> {
            String message = new String(delivery.getBody(), "UTF-8");
            System.out.println(" [x] Received '" + message + "'");
            try {
                // Simulate long-running task
                Thread.sleep(1000); // 1 second per task
            } catch (InterruptedException e) {
                Thread.currentThread().interrupt();
            } finally {
                System.out.println(" [x] Done with '" + message + "'");
            }
        };
        channel.basicConsume(QUEUE_NAME, true, deliverCallback, consumerTag -> {});
    }
}

Consumer Code Breakdown

Let's look at the consumer's logic:

Similar to the producer, it connects and declares the queue.
A DeliverCallback defines the actions when a message arrives.
Inside the callback, Thread.sleep(1000) simulates a 1-second task.
channel.basicConsume(QUEUE_NAME, true, deliverCallback, ...) starts consuming. The true means messages are automatically acknowledged after delivery.

Scale with Multiple Workers

Here's the core demonstration of work queues:

1. Run the TaskConsumer code in two separate terminal windows (or instances).

2. Then, run the TaskProducer code once.

You will observe that the 10 tasks are divided between your two worker instances, each processing roughly 5 tasks due to RabbitMQ's round-robin dispatch.

Work Queue Quiz

Imagine you have a single RabbitMQ queue and two consumers (Worker A and Worker B) listening to it. A producer sends 4 messages (M1, M2, M3, M4) to this queue.

Which statement accurately describes how the messages are typically distributed using RabbitMQ's default fair dispatch?

Work Queues: Key Takeaways

You've successfully implemented Work Queues, a fundamental pattern for distributing tasks across multiple consumers!

Work queues enable asynchronous processing and prevent single points of failure.
RabbitMQ's default round-robin dispatch strategy ensures tasks are distributed fairly.
By running multiple consumer instances, you can easily scale your task processing capacity.

Next, we'll dive into making your message handling even more robust with acknowledgements and message durability!

よくある質問

「Work Queues：公平なディスパッチ」レッスンは無料ですか？

はい。「Work Queues：公平なディスパッチ」の完全なテキストはこのウェブで無料で読めます。インタラクティブに演習し（組み込みコードエディタと24時間対応のAIチューター）、RabbitMQ Messaging & Async Systemsコースの残りをアンロックするには、CoddyKit PROにアップグレードしてください。 RabbitMQ Messaging & Async Systemsコースには全4レッスンが含まれています。

「Work Queues：公平なディスパッチ」で何を学びますか？

ラウンドロビン方式で複数の Consumer にタスクを分配する Work Queues を実装します。時間のかかるタスクを非同期に処理する方法を学びます。ブラウザで直接実行するハンズオンコードでRabbitMQ Messaging & Async Systemsを演習し、24時間対応のAIチューターがレッスンを進める中での質問に答えます。

RabbitMQ Messaging & Async Systemsを始めるのに経験は必要ですか？

事前経験は必要ありません。CoddyKitのRabbitMQ Messaging & Async Systemsは初級者から上級者向けに構成されているため、ここから始めるか最初から始めて、自分のペースで進むことができます。これはレッスン2/4です。

「Work Queues：公平なディスパッチ」レッスンにはどのくらい時間がかかりますか？

ほとんどのCoddyKitレッスンは約5～10分かかります。各レッスンはコンパクトでインタラクティブなので、着実に進歩し、ウェブとアプリ全体で正確に前回の場所から再開できます。

このRabbitMQ Messaging & Async Systemsレッスンでコードを書いて実行できますか？

はい。すべてのRabbitMQ Messaging & Async Systemsレッスンに組み込みコードエディタが含まれているため、ブラウザでリアルコードを書いて実行し、即座のAIフィードバックを取得できます。ローカル設定は不要です。

このコースのすべてのレッスン

← RabbitMQ Messaging & Async Systemsに戻る