平常项目中很多场景需要使用延时消息处理,例如订单超过多久没有支付需要取消等。如何在消息中间件RabbitMQ中实现该功能?下面描述下使用
Dead Letter Exchange
实现延时消息场景,当然会有别的其他实现方式。
1. 什么是Dead Letter Exchange?
RabbitMQ中通常消息被直接发送到队列中或者从Exchange中Route到队列上后,消息如果被消费者消费完毕并确认后消息就会从Broker中被删除。
如果存在以下三种情况,同时队列上设置了Dead Letter Exchange
,消息会被转送到Dead Letter Exchange
中。 消息被拒绝(
basicReject
或者basicNack
) requeue=false消息存活时间超过了TTL预设值(
x-message-ttl
)队列满了
Dead Letter Exchange像平常的Exchange一样,可以设置它的BuiltinExchangeType
,也可以为它绑定队列。
2. 代码实例
流程先
I. 定义消息生产者
/*** * 消息发送者 */static class NormalEXSend { private Connection conn; private Channel chnl; public NormalEXSend(String tag) throws IOException, TimeoutException { ConnectionFactory connFact = initConnFac(); conn = connFact.newConnection(); chnl = conn.createChannel(); // 定义正常工作Exchange chnl.exchangeDeclare(WORKER_EXCHANGE_NAME, BuiltinExchangeType.DIRECT); // 定义 dead letter exchange chnl.exchangeDeclare(DELAY_EXCHANGE_NAME, BuiltinExchangeType.DIRECT); Mapargs = new HashMap<>(); args.put("x-message-ttl", 60000); // timeout 1min args.put("x-dead-letter-exchange", DELAY_EXCHANGE_NAME); args.put("x-dead-letter-routing-key", DEAD_ROUTING_KEY); // 定义正常工作Queue同时设置dead letter exchange chnl.queueDeclare(WORKER_QUEUE_NAME, false, false, false, args); // 绑定到正常工作Exchange chnl.queueBind(WORKER_QUEUE_NAME, WORKER_EXCHANGE_NAME, tag); } /** * 发送消息 * @param key * @param msg * @throws IOException */ public void send(String key, String msg) throws IOException { AMQP.BasicProperties props = MessageProperties.PERSISTENT_TEXT_PLAIN; // send a message to a exchange chnl.basicPublish(WORKER_EXCHANGE_NAME, key, props, msg.getBytes()); System.out.println(String.format("[%s|%s|Sender] send 【%s】 to exchange:%s", Thread.currentThread().getName(), System.currentTimeMillis(), msg, WORKER_EXCHANGE_NAME)); }}
II. 定义延时消息处理者
其中receive方法中
consumer
的handleDelivery
方法参数properties
可以获取到消息的death原因properties.getHeaders().get("x-first-death-reason")
,可能值rejected | expired | maxlen
。此处可以根据判断此值去处理由于超时而引起death的消息(就是我们想要处理的延时消息)。
/** * 延时消息处理者 */ static class DelayEXRecv { private Connection conn; private Channel chnl; public DelayEXRecv() throws IOException, TimeoutException { ConnectionFactory connFact = initConnFac(); conn = connFact.newConnection(); chnl = conn.createChannel(); // 定义延时消息队列 chnl.queueDeclare(DELAY_QUEUE_NAME, false, false, false, null); // 绑定到延时消息Exchange chnl.queueBind(DELAY_QUEUE_NAME, DELAY_EXCHANGE_NAME, DEAD_ROUTING_KEY); } /** * 接收消息 * @throws IOException */ public void receive() throws IOException { chnl.basicQos(1); // no auto ack boolean autoAck = false; chnl.basicConsume(DELAY_QUEUE_NAME, autoAck, new DefaultConsumer(chnl) { public void handleDelivery(String consumerTag, Envelope envelope, AMQP.BasicProperties properties, byte[] body) throws IOException { String message = new String(body, "UTF-8"); // 打印出延时原因 rejected | expired | maxlen // 项目中可以根据原因处理目标消息 System.out.println(String.format("[%s|%s|Delay_Receiver] received the delay msg 【%s】 from EXCHANGE: %s, the delay reason is: %s", Thread.currentThread().getName(), System.currentTimeMillis(), message, envelope.getExchange(), properties.getHeaders().get("x-first-death-reason"))); // 确认消息 chnl.basicAck(envelope.getDeliveryTag(), false); } }); } }
III. 试验一把
private static final String WORKER_EXCHANGE_NAME = "exchange.worker"; private static final String DELAY_EXCHANGE_NAME = "exchange.delay"; private static final String WORKER_QUEUE_NAME = "queue.worker"; private static final String DELAY_QUEUE_NAME = "queue.delay"; private static final String DEAD_ROUTING_KEY = "dead.key.message"; public static void main(String[] args) { ExecutorService exec = Executors.newFixedThreadPool(2); exec.execute(new Runnable() { @Override public void run() { try { String key = "worker"; NormalEXSend sender = new NormalEXSend(key); for (int i =0; i < 5; i++) { sender.send(key, String.format("YaYYY, one message, No.:%s!", i)); Thread.sleep(3000); } } catch (IOException | TimeoutException | InterruptedException e) { e.printStackTrace(); } } }); exec.execute(new Runnable() { @Override public void run() { try { DelayEXRecv receiver = new DelayEXRecv(); receiver.receive(); System.out.println(String.format("[%s|%s|Delay_Receiver] Starting the Delay Msg Receiver process...", Thread.currentThread().getName(), System.currentTimeMillis())); } catch (IOException | TimeoutException e) { e.printStackTrace(); } } }); exec.shutdown(); }
IV. 打印结果
[pool-1-thread-2|1515750089010|Delay_Receiver] Starting the Delay Msg Receiver process...[pool-1-thread-1|1515750089020|Sender] send 【YaYYY, one message, No.:0!】 to exchange:exchange.worker[pool-1-thread-1|1515750092020|Sender] send 【YaYYY, one message, No.:1!】 to exchange:exchange.worker[pool-1-thread-1|1515750095020|Sender] send 【YaYYY, one message, No.:2!】 to exchange:exchange.worker[pool-1-thread-1|1515750098021|Sender] send 【YaYYY, one message, No.:3!】 to exchange:exchange.worker[pool-1-thread-1|1515750101022|Sender] send 【YaYYY, one message, No.:4!】 to exchange:exchange.worker[pool-2-thread-4|1515750149038|Delay_Receiver] received the delay msg 【YaYYY, one message, No.:0!】 from EXCHANGE: exchange.delay, the delay reason is: expired[pool-2-thread-5|1515750152035|Delay_Receiver] received the delay msg 【YaYYY, one message, No.:1!】 from EXCHANGE: exchange.delay, the delay reason is: expired[pool-2-thread-6|1515750155035|Delay_Receiver] received the delay msg 【YaYYY, one message, No.:2!】 from EXCHANGE: exchange.delay, the delay reason is: expired[pool-2-thread-7|1515750158036|Delay_Receiver] received the delay msg 【YaYYY, one message, No.:3!】 from EXCHANGE: exchange.delay, the delay reason is: expired[pool-2-thread-8|1515750161036|Delay_Receiver] received the delay msg 【YaYYY, one message, No.:4!】 from EXCHANGE: exchange.delay, the delay reason is: expired
可以看出消息是在制定延时的1min后才被获取消费。
Yayy, 至此结束。参考:http://www.rabbitmq.com/dlx.html