Work queues 工作队列模式
模式说明
通过Helloworld工程我们已经能够构建一个简单的消息队列的基本项目,项目中存在几个角色:生产 者、消费者、队列,而对于我们真实的开发中 ,对于消息的消费者通过是有多个的。
比如在实现用户注册功能时,用户注册成功,会给响对应用户发送邮件,同时给用户发送手机短信,告诉用户已成功注册网站或者app 应用,这种功能在大部分项目开发中都比较常见 ,而对于helloworld 的应用中虽然能够对 消息进行消费,但是有很大问题: 消息消费者只有一个,当消息量非常大时,单个消费者处理消息就会变得很慢,同时给节点页带来很大压力,导致消息堆积越来越多。对于这种情况,RabbitMQ 提供了工作 队列模式,通过工作队列提供做个消费者,对MQ产生的消息进行消费,提高MQ消息的吞吐率,降低消息的处理时间。处理模型图如下。
实现步骤
生产者
package cn.wolfcode.java.rabbitmq._02worker;
import com.rabbitmq.client.Channel;
import com.rabbitmq.client.Connection;
import com.rabbitmq.client.ConnectionFactory;
import com.rabbitmq.client.MessageProperties;/*** Created by wolfcode-fanjialong*/public class NewTask {private static final String TASK_QUEUE_NAME = "task_queue";public static void main(String[] argv) throws Exception {ConnectionFactory factory = new ConnectionFactory();factory.setHost("192.168.142.129");try (Connection connection = factory.newConnection();Channel channel = connection.createChannel()) {channel.queueDeclare(TASK_QUEUE_NAME, true, false, false, null);String message = "tasequeue";for(int i=0;i<20;i++){channel.basicPublish("", TASK_QUEUE_NAME,MessageProperties.PERSISTENT_TEXT_PLAIN,(message+i).getBytes("UTF-8"));}System.out.println(" [x] Sent '" + message + "'");}}
}
消费者
package cn.wolfcode.java.rabbitmq._02worker;import com.rabbitmq.client.Channel;
import com.rabbitmq.client.Connection;
import com.rabbitmq.client.ConnectionFactory;
import com.rabbitmq.client.DeliverCallback;
import java.util.concurrent.TimeUnit;/*** Created by wolfcode-fanjialong*/public class Worker {private static final String TASK_QUEUE_NAME = "task_queue";public static void main(String[] argv) throws Exception {ConnectionFactory factory = new ConnectionFactory();factory.setHost("192.168.142.129");final Connection connection = factory.newConnection;final Channel channel = connection.createChannel();channel.queueDeclare(TASK_QUEUE_NAME, true, false, false, null);System.out.println(" [*] Waiting for messages. To exit press CTRL+C");//channel.basicQos(1);DeliverCallback deliverCallback = (consumerTag, delivery) -> {String message = new String(delivery.getBody(), "UTF-8");try {TimeUnit.SECONDS.sleep(1);} catch (InterruptedException e) {e.printStackTrace();}finally {channel.basicAck(delivery.getEnvelope().getDeliveryTag(), false);}System.out.println(" [x] Received '" + message + "'");};channel.basicConsume(TASK_QUEUE_NAME, false, deliverCallback, consumerTag -> { });}}
简单问题说明
从结果可以看出消息被平均分配到两个消费方,来对消息进行处理,提高了消息处理效率,创建多个消费者来对消息进行处理。这里RabitMQ采用轮询来对消息进行分发时保证了消息被平均分配到每个消费方 。
但是引入新的问题:真正的生产环境下,对于消息的处理基本不会像我们现在看到的这样,每个消 费方处理的消息数量是平均分配的,比如因为网络原因,机器cpu ,内存等硬问题,消费方处理消息时 同类消息不同机器进行处理时消耗时间也是不一样的,比如1号消费者消费1条消息时1秒,2号消费者消费1条消息是5秒,对于1号消费者比2号消费者处理消息快,那么在分配消息时就应该让1号消费者多收 到消息进行处理,也即是我们通常所说的”能者多劳”,同样Rabbitmq对于这种消息分配模式提供了支持。
问题: 任务量很大,消息虽然得到了及时的消费,单位时间内消息处理速度加快,提高了吞吐量,可 是不同消费者处理消息的时间不同,导致部分消费者的资源被浪费。
解决:采用消息公平分发。
总结:工作队列消息轮询分发消费者收到的消息数量平均分配,单位时间内消息处理速度加快,提高了吞吐量。
工作模式队列-消息公平分发(fair dispatch)
在案例01中对于消息分发采用的是默认轮询分发,消息应答采用的自动应答模式,这是因为当消息进 入队列,RabbitMQ就会分派消息。它不看消费者为应答的数目,只是盲目的将第n条消息发给第n个消费者。
为了解决这个问题,我们使用 basicQos(prefetchCount = 1) 方法,来限RabbitMQ只发不超过1条的消息给同一个消费者。当消息处理完毕后,有了反馈,才会进行第二次发送。执行模型图如下:
Pub/Sub 订阅模式
模式说明
在订阅模型中,多了一个 Exchange 角色,而且过程略有变化:
P:生产者,也就是要发送消息的程序,但是不再发送到队列中,而是发给X(交换机)
C:消费者,消息的接收者,会一直等待消息到来
Queue:消息队列,接收消息、缓存消息
Exchange:交换机(X)。一方面,接收生产者发送的消息。另一方面,知道如何处理消息,例如递交给某个特别队列、
递交给所有队列、或是将消息丢弃。到底如何操作,取决于Exchange的类型。Exchange有常见以下3种类型:
-
Fanout:广播,将消息交给所有绑定到交换机的队列
-
Direct:定向,把消息交给符合指定routing key 的队列
-
Topic:通配符,把消息交给符合routing pattern(路由模式) 的队列
Exchange(交换机)只负责转发消息,不具备存储消息的能力,因此如果没有任何队列与 Exchange 绑定,或者没有符合路由规则的队列,那么消息会丢失!
实现步骤
生产者
package cn.wolfcode.java.rabbitmq._03pubsub;
import com.rabbitmq.client.Channel;
import com.rabbitmq.client.Connection;
import com.rabbitmq.client.ConnectionFactory;/*** Created by wolfcode-fanjialong*/
public class EmitLog {private static final String EXCHANGE_NAME = "logs";public static void main(String[] argv) throws Exception {ConnectionFactory factory = new ConnectionFactory();factory.setHost("localhost");try (Connection connection = factory.newConnection();Channel channel = connection.createChannel()) {channel.exchangeDeclare(EXCHANGE_NAME, "fanout");String message = "info: Hello World!";channel.basicPublish(EXCHANGE_NAME, "", null, message.getBytes("UTF-8"));System.out.println(" [x] Sent '" + message + "'");}}
}
消费者
package cn.wolfcode.java.rabbitmq._03pubsub;import com.rabbitmq.client.Channel;
import com.rabbitmq.client.Connection;
import com.rabbitmq.client.ConnectionFactory;
import com.rabbitmq.client.DeliverCallback;/*** Created by wolfcode-fanjialong*/public class ReceiveLogs {private static final String EXCHANGE_NAME = "logs";public static void main(String[] argv) throws Exception {ConnectionFactory factory = new ConnectionFactory();factory.setHost("localhost");Connection connection = factory.newConnection();Channel channel = connection.createChannel();channel.exchangeDeclare(EXCHANGE_NAME, "fanout");String queueName = channel.queueDeclare().getQueue();channel.queueBind(queueName, EXCHANGE_NAME, "");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 + "'");};channel.basicConsume(queueName, true, deliverCallback, consumerTag -> { });}
}
Routing 路由模式
模式说明
队列与交换机的绑定,不能是任意绑定了,而是要指定一个 RoutingKey(路由key)
消息的发送方在向 Exchange 发送消息时,也必须指定消息的 RoutingKey
Exchange 不再把消息交给每一个绑定的队列,而是根据消息的 Routing Key 进行判断,只有队列的 Routingkey 与消息的 Routing key 完全一致,才会接收到消息
图解:
P:生产者,向 Exchange 发送消息,发送消息时,会指定一个routing key
X:Exchange(交换机),接收生产者的消息,然后把消息递交给与 routing key 完全匹配的队列
C1:消费者,其所在队列指定了需要 routing key 为 error 的消息
C2:消费者,其所在队列指定了需要 routing key 为 info、error、warning 的消息
实现步骤
生产者
package cn.wolfcode.java.rabbitmq._04rounting;
import com.rabbitmq.client.BuiltinExchangeType;
import com.rabbitmq.client.Channel;
import com.rabbitmq.client.Connection;
import com.rabbitmq.client.ConnectionF![在这里插入图片描述](https://img-blog.csdnimg.cn/direct/6bffe712644346098952dba200523856.png)
actory;
/*** Created by wolfcode-fanjialong*/public class EmitLogDirect {private static final String EXCHANGE_NAME = "direct_logs";public static void main(String[] argv) throws Exception {ConnectionFactory factory = new ConnectionFactory();factory.setHost("localhost");try (Connection connection = factory.newConnection();Channel channel = connection.createChannel()) {channel.exchangeDeclare(EXCHANGE_NAME, BuiltinExchangeType.DIRECT);String severity = "info";String message = "directMsg";channel.basicPublish(EXCHANGE_NAME, severity, null, message.getBytes("UTF-8"));System.out.println(" [x] Sent '" + severity + "':'" + message + "'");}}
}
消费者
package cn.wolfcode.java.rabbitmq._04rounting;import com.rabbitmq.client.*;/*** Created by wolfcode-fanjialong*/public class ReceiveLogsDirect {private static final String EXCHANGE_NAME = "direct_logs";public static void main(String[] argv) throws Exception {ConnectionFactory factory = new ConnectionFactory();factory.setHost("localhost");Connection connection = factory.newConnection();Channel channel = connection.createChannel();channel.exchangeDeclare(EXCHANGE_NAME, BuiltinExchangeType.DIRECT);String queueName = channel.queueDeclare().getQueue();channel.queueBind(queueName, EXCHANGE_NAME, "info");channel.queueBind(queueName, EXCHANGE_NAME, "error");channel.queueBind(queueName, EXCHANGE_NAME, "warning");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 '" + delivery.getEnvelope().getRoutingKey() + "':'" + message + "'");};channel.basicConsume(queueName, true, deliverCallback, consumerTag -> {});}}
Topic 模式
模式介绍
Topic 类型与 Direct 相比,都是可以根据 RoutingKey 把消息路由到不同的队列。只不过 Topic 类型
Exchange 可以让队列在绑定 Routing key 的时候使用通配符!
Routingkey 一般都是有一个或多个单词组成,多个单词之间以”.”分割,例如: item.insert
通配符规则:# 匹配一个或多个词,* 匹配不多不少恰好1个词,例如:item.# 能够匹配 item.insert.abc 或者 item.insert,item.* 只能匹配 item.insert
图解:
红色 Queue:绑定的是 usa.# ,因此凡是以 usa. 开头的 routing key 都会被匹配到
黄色 Queue:绑定的是 #.news ,因此凡是以 .news 结尾的 routing key 都会被匹配
实现步骤
生产者
package cn.wolfcode.java.rabbitmq._05topic;import com.rabbitmq.client.Channel;
import com.rabbitmq.client.Connection;
import com.rabbitmq.client.ConnectionFactory;/*** Created by wolfcode-fanjialong*/public class EmitLogTopic {private static final String EXCHANGE_NAME = "topic_logs";public static void main(String[] argv) throws Exception {ConnectionFactory factory = new ConnectionFactory();factory.setHost("localhost");try (Connection connection = factory.newConnection();Channel channel = connection.createChannel()) {channel.exchangeDeclare(EXCHANGE_NAME, "topic");String routingKey = "order1.save";String message = "topicMsg";channel.basicPublish(EXCHANGE_NAME, routingKey, null, message.getBytes("UTF-8"));System.out.println(" [x] Sent '" + routingKey + "':'" + message + "'");}}}
消费者
package cn.wolfcode.java.rabbitmq._05topic;import com.rabbitmq.client.Channel;
import com.rabbitmq.client.Connection;
import com.rabbitmq.client.ConnectionFactory;
import com.rabbitmq.client.DeliverCallback;/*** Created by wolfcode-fanjialong*/public class ReceiveLogsTopic {private static final String EXCHANGE_NAME = "topic_logs";public static void main(String[] argv) throws Exception {ConnectionFactory factory = new ConnectionFactory();factory.setHost("localhost");Connection connection = factory.newConnection();Channel channel = connection.createChannel();channel.exchangeDeclare(EXCHANGE_NAME, "topic");String queueName = channel.queueDeclare().getQueue();channel.queueBind(queueName, EXCHANGE_NAME, "order.*");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 '" + delivery.getEnvelope().getRoutingKey() + "':'" + message + "'");};channel.basicConsume(queueName, true, deliverCallback, consumerTag -> { });}}