RabbitMQ Messaging & Async Systems · 课时

预取数量（QoS）

配置预取数量（服务质量），控制消费者一次接收的消息数量。优化消费者效率，避免单个消费者不堪重负。

预取数量（QoS） 是 CoddyKit 上的免费 RabbitMQ Messaging & Async Systems 课时。这是第 2 节课，共 4 节。你可以在下方免费阅读本课时的完整内容 — 然后在浏览器中使用内置代码编辑器和全天候 AI 导师进行实践。这是 RabbitMQ Messaging & Async Systems 学习路径的一部分，你的进度在网页和 CoddyKit 应用中同步。 RabbitMQ Messaging & Async Systems 课程共包含 4 节课。

本课时的部分内容尚未翻译，以英文显示。

Control Message Flow with Prefetch

In a system with multiple consumers, how do you ensure messages are distributed fairly and no single consumer gets overwhelmed?

This lesson introduces the prefetch count, a crucial Quality of Service (QoS) setting in RabbitMQ.

The Challenge: Uneven Workload

By default, RabbitMQ dispatches messages in a round-robin fashion to available consumers. However, it doesn't wait for a consumer to finish processing a message before sending the next one.

If one consumer is slow, it might still receive many messages, while a fast consumer sits idle. This leads to an uneven workload.

Introducing Quality of Service (QoS)

RabbitMQ's Quality of Service (QoS) settings allow you to control how messages are delivered to consumers.

The most common QoS setting is the prefetchCount, which limits the number of unacknowledged messages a consumer can hold at any given time.

How Prefetch Count Works

The prefetch count tells RabbitMQ the maximum number of messages a consumer is willing to process at once.

Once a consumer reaches its prefetch limit, RabbitMQ stops delivering new messages to it.
It will only send more messages once the consumer acknowledges some of its current messages, bringing the unacknowledged count below the limit.

Implementing Prefetch with `basicQos`

You set the prefetch count using the channel.basicQos() method on the consumer side. The most common value for fair dispatch is 1.

This means a consumer will process one message at a time, acknowledge it, then receive the next.

    // Consumer setup
    Channel channel = connection.createChannel();
    channel.queueDeclare("my_queue", true, false, false, null);

    // Set prefetch count to 1
    channel.basicQos(1, false); // prefetchCount = 1, global = false

    DeliverCallback deliverCallback = (consumerTag, delivery) -> {
        String message = new String(delivery.getBody(), "UTF-8");
        System.out.println(" [x] Received '" + message + "'");
        // Simulate work
        Thread.sleep(1000);
        System.out.println(" [x] Done processing '" + message + "'");
        channel.basicAck(delivery.getEnvelope().getDeliveryTag(), false);
    };
    channel.basicConsume("my_queue", false, deliverCallback, consumerTag -> {});

Sending Messages to the Queue

Let's create a simple producer that sends several messages to a queue. We'll use this with our QoS-enabled consumer.

Run this code first to populate the queue:

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

public class MessageProducer {
    private final static String QUEUE_NAME = "my_queue";

    public static void main(String[] argv) throws Exception {
        ConnectionFactory factory = new ConnectionFactory();
        factory.setHost("localhost"); // Assuming RabbitMQ is on localhost
        try (Connection connection = factory.newConnection();
             Channel channel = connection.createChannel()) {
            channel.queueDeclare(QUEUE_NAME, true, false, false, null);

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

Consumer with Prefetch in Action

Now, run this consumer code. Notice how basicQos(1) ensures that even if a consumer is slow (due to Thread.sleep), it won't hoard messages.

If you run multiple instances of this consumer, they will share the work more fairly.

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

public class QosConsumer {
    private final static String QUEUE_NAME = "my_queue";

    public static void main(String[] argv) throws Exception {
        ConnectionFactory factory = new ConnectionFactory();
        factory.setHost("localhost"); // Assuming RabbitMQ is on localhost
        Connection connection = factory.newConnection();
        Channel channel = connection.createChannel();

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

        // Set prefetch count to 1 for fair dispatch
        channel.basicQos(1, false);

        DeliverCallback deliverCallback = (consumerTag, delivery) -> {
            String message = new String(delivery.getBody(), "UTF-8");
            System.out.println(" [x] Received '" + message + "'");
            try {
                // Simulate processing time
                Thread.sleep(2000); // 2 seconds per message
            } catch (InterruptedException e) {
                Thread.currentThread().interrupt();
            } finally {
                System.out.println(" [x] Done processing '" + message + "'");
                channel.basicAck(delivery.getEnvelope().getDeliveryTag(), false);
            }
        };
        channel.basicConsume(QUEUE_NAME, false, deliverCallback, consumerTag -> {});
    }
}

Key Advantages of Using Prefetch

Using an appropriate prefetch count offers several benefits:

Fair Dispatch: Messages are distributed more evenly among competing consumers, preventing a single consumer from getting overloaded.
Prevents Consumer Overload: Slower consumers won't accumulate too many messages, reducing memory consumption and potential crashes.
Better Resource Utilization: Ensures all active consumers are working efficiently, leading to better overall throughput for the system.

`global` Flag for Channel-Wide QoS

The basicQos() method has an optional global parameter. If set to true, the prefetch count applies to all consumers on that channel.

Generally, it's safer to set global to false (the default) so the prefetch count applies per-consumer. This offers finer control and prevents unexpected behavior.

    // Per-consumer QoS (recommended)
    channel.basicQos(1, false);

    // Channel-wide QoS (use with caution)
    // channel.basicQos(1, true);

Optimizing Your Prefetch Count

The ideal prefetch count depends on your application's specifics. Consider:

Message Processing Time: If messages are processed quickly, a higher prefetch count can increase throughput.
Consumer Resources: How much memory and CPU can each consumer dedicate to holding and processing messages?
Network Latency: For high-latency networks, a slightly higher prefetch can reduce idle time waiting for the next message.

Experimentation is key to finding the optimal balance between throughput and fairness.

Prefetch Count Understanding

Consider two consumers, C1 and C2, both connected to the same queue. C1 processes messages in 5 seconds, C2 in 1 second. If basicQos(1) is set for both, and 10 messages are sent, what is the primary benefit?

Recap: Controlling Message Flow

Great job! You've learned about the prefetch count (QoS) in RabbitMQ.

It limits unacknowledged messages a consumer holds.
channel.basicQos(1) is common for fair dispatch.
It prevents consumer overload and ensures even workload distribution.

Mastering prefetch count is vital for building robust and scalable messaging systems with competing consumers.

常见问题解答

「预取数量（QoS）」课时是免费的吗？

是的 — 「预取数量（QoS）」的完整文本可在网页上免费阅读。要进行交互式练习（内置代码编辑器和全天候 AI 导师）并解锁 RabbitMQ Messaging & Async Systems 课程的其余内容，请升级到 CoddyKit PRO。 RabbitMQ Messaging & Async Systems 课程共包含 4 节课。

「预取数量（QoS）」这节课中我会学到什么？

配置预取数量（服务质量），控制消费者一次接收的消息数量。优化消费者效率，避免单个消费者不堪重负。你通过在浏览器中直接运行的动手代码来练习 RabbitMQ Messaging & Async Systems，全天候 AI 导师会在你学习这节课的过程中回答你的问题。

学习 RabbitMQ Messaging & Async Systems 需要有经验吗？

无需任何先前经验。CoddyKit 上的 RabbitMQ Messaging & Async Systems 课程适合初学者到高级学习者，你可以从这里开始或从头开始，按照自己的节奏学习。这是第 2 节课，共 4 节。

「预取数量（QoS）」课时需要多长时间？

大多数 CoddyKit 课程大约需要 5–10 分钟。每节课都很精短且互动，所以你能稳步进步，并在网页和应用中从离开的地方继续。

我能在这节 RabbitMQ Messaging & Async Systems 课中编写并运行代码吗？

能。每节 RabbitMQ Messaging & Async Systems 课都包含内置代码编辑器，你可以在浏览器中直接编写并运行真实代码，并获得即时 AI 反馈 — 无需本地设置。

此课程中的所有课时

← 返回 RabbitMQ Messaging & Async Systems