HA向けミラーキュー
複数のクラスター・ノード間でメッセージを複製するミラーキューを実装します。ノード障害が発生してもメッセージが失われない、高可用性のキューを実現します。
「HA向けミラーキュー」はCoddyKit上の無料RabbitMQ Messaging & Async Systemsレッスンです。 これはレッスン3/4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。
Intro to Mirrored Queues
In a RabbitMQ cluster, queues are by default located on a single node. If that node fails, any messages in its queues (and the queues themselves) become unavailable until the node recovers.
Mirrored queues solve this by replicating queue contents across multiple nodes. This ensures high availability (HA) and fault tolerance for your messages.
Why Use Mirrored Queues?
Imagine a critical application where losing messages or experiencing downtime is unacceptable. Mirrored queues provide:
- High Availability: If the node hosting the primary queue fails, a replica can take over seamlessly.
- Data Durability: Messages are stored on multiple nodes, protecting against single-node failures.
- Fault Tolerance: The system can continue operating even if some nodes go offline.
They are essential for robust, production-grade RabbitMQ deployments.
Master & Replica Architecture
When a queue is mirrored, one node hosts the master (or primary) queue. All other nodes hosting mirrors run replicas.
All operations for a mirrored queue (publishing, consuming, adding messages) are first handled by the master. The master then replicates these operations to all its replicas.
Configuring Mirroring with Policies
You don't configure mirroring directly on individual queues. Instead, you use policies. A policy is a set of rules that apply to queues whose names match a specific pattern.
This allows you to define mirroring behavior for many queues at once, or for queues created in the future, without modifying client code.
Policy Example: Mirror All Queues
Here's how to create a policy named ha-all that mirrors all queues (matching ".*") to all nodes in the cluster ("ha-mode":"all").
You'd typically run this command on one of your RabbitMQ cluster nodes via the command line.
rabbitmqctl set_policy ha-all ".*" '{"ha-mode":"all"}' --apply-to queuesUnderstanding ha-mode Options
The ha-mode argument in a policy defines how mirroring should behave:
all: The queue will be mirrored to all nodes in the cluster.exactly: The queue will be mirrored to a specific number of nodes (e.g.,{"ha-mode":"exactly", "ha-params":2}).nodes: The queue will be mirrored to a specific list of named nodes (e.g.,{"ha-mode":"nodes", "ha-params":["rabbit@node1", "rabbit@node2"]}).
Producers & Mirrored Queues
From a producer's perspective, interacting with a mirrored queue is no different than a non-mirrored one. The client connects to any node in the cluster, and RabbitMQ handles routing the message to the master queue for mirroring.
Try running this simple Java producer:
import com.rabbitmq.client.Channel;
import com.rabbitmq.client.Connection;
import com.rabbitmq.client.ConnectionFactory;
public class Producer {
private final static String QUEUE_NAME = "my_mirrored_queue";
public static void main(String[] argv) throws Exception {
ConnectionFactory factory = new ConnectionFactory();
factory.setHost("localhost"); // Connect to any cluster node
try (Connection connection = factory.newConnection();
Channel channel = connection.createChannel()) {
// Declare a durable queue (important for mirrored queues)
channel.queueDeclare(QUEUE_NAME, true, false, false, null);
String message = "Hello, Mirrored Queue!";
channel.basicPublish("", QUEUE_NAME, null, message.getBytes("UTF-8"));
System.out.println(" [x] Sent '" + message + "'");
}
}
}Consumers & Mirrored Queues
Similarly, consumers don't need special logic to consume from a mirrored queue. They simply connect to a node and subscribe to the queue.
If the master queue fails, RabbitMQ automatically promotes a replica to master, and consumers transparently switch to the new master (though a brief reconnect might be needed).
Run this consumer example:
import com.rabbitmq.client.Channel;
import com.rabbitmq.client.Connection;
import com.rabbitmq.client.ConnectionFactory;
import com.rabbitmq.client.DeliverCallback;
public class Consumer {
private final static String QUEUE_NAME = "my_mirrored_queue";
public static void main(String[] argv) throws Exception {
ConnectionFactory factory = new ConnectionFactory();
factory.setHost("localhost"); // Connect to any cluster node
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");
DeliverCallback deliverCallback = (consumerTag, delivery) -> {
String message = new String(delivery.getBody(), "UTF-8");
System.out.println(" [x] Received '" + message + "'");
};
// Basic consume with auto-acknowledgement
channel.basicConsume(QUEUE_NAME, true, deliverCallback, consumerTag -> {});
}
}Failover & Replica Sync
When a master node fails, RabbitMQ elects a new master from the available replicas. This new master takes over, and message processing continues without data loss.
If a node with a replica rejoins the cluster, or a new node is added, the replica will synchronize its contents with the current master. This ensures all messages are consistent across the mirrored queue instances.
Quick Check: Mirrored Queues
You've learned about the importance and mechanics of mirrored queues. Let's test your understanding.
Recap: Mirrored Queues for HA
We've explored mirrored queues, a crucial feature for ensuring high availability and data durability in RabbitMQ clusters.
- Mirrored queues replicate messages across master and replica nodes.
- They are configured using policies, allowing flexible control over mirroring behavior.
- Client applications (producers/consumers) interact with mirrored queues transparently.
- In case of a master node failure, a replica is promoted, ensuring continuous service and message safety.
This mechanism is vital for building robust, fault-tolerant messaging systems.
よくある質問
「HA向けミラーキュー」レッスンは無料ですか？
はい。「HA向けミラーキュー」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。
「HA向けミラーキュー」で何を学びますか？
複数のクラスター・ノード間でメッセージを複製するミラーキューを実装します。ノード障害が発生してもメッセージが失われない、高可用性のキューを実現します。 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3/4です。
「HA向けミラーキュー」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。