开发者

RocketMQ根据Tag进行消息过滤

开发者 https://www.devze.com 2023-02-14 10:39 出处:网络 作者: 每天都要进步一点点
目录一、概述二、Tag过滤三、根据自定义属性进行过滤 (SQL过滤)一、概述 RocketMQ的消费者可以根据Tag进行消息过滤,也支持自定义属性过滤。消息过滤目前是在Broker端实现的,优点是减少了对于Consumer无用消息的网络
目录
  • 一、概述
  • 二、Tag过滤
  • 三、根据自定义属性进行过滤 (SQL过滤)

一、概述

RocketMQ的消费者可以根据Tag进行消息过滤,也支持自定义属性过滤。消息过滤目前是在Broker端实现的,优点是减少了对于Consumer无用消息的网络传输,缺点是增加了Broker的负担、而且实现相对复杂。RocketMQ支持两种方式的消息过滤。一种是Tag过滤,另外一种是SQL过滤。下面我们分别介绍一下。

二、Tag过滤

在大多数情况下,Tag是个简单而有用的设计,其可以来选择您想要的消息。下面我们通过一个示例演示:

(1)、生产者发送消息

public class MQProducer {
    public static void main(String[] args) throws MQClientException, UnsupportedEncodingException, RemotingException, InterruptedException, MQBrokerException {
        // 创建DefaultMQProducer类并设定生产者名称
        DefaultMQProducer mqProducer = new DefaultMQProducer("producer-group-test");
        // 设置NameServer地址,如果是集群的话,使用分号;分隔开
        mqProducer.setNamesrvAddr("10.0.90.86:9876");
        // 消息最大长度 默认4M
        mqProducer.setMaxMessageSize(4096);
        // 发送消息超时时间,默认3000
        mqProducer.setSendMsgTimeout(3000);
        // 发送消息失败重试次数,默认2
        mqProducer.setRetryTimesWhenSendAsyncFailed(2);
        // 启动消息生产者
        mqProducer.start();
        String[] tags = new String[]{"TagA", "TagB", "TagC", "TagD", "TagE"};
        for (int i = 0; i < 10; i++) {
            String tag = tags[i % tags.length];
            String msg = "hello, 这是第" + (i + 1) + "条消息";
            // 创建消息,并指定Topic(主题),Tag(标签)和消息内容
            Message message = new Message("FilterMessageTopic", tag, msg.getBytes(RemotingHelper.DEFAULT_CHARSET));
            // 发送同步消息到一个Broker,可以通过sendResult返回消息是否成功送达
            SendResult sendResult = mqProducer.send(message);
            System.out.println(sendResult);
        }
        // 如果不再发送消息,关闭Producer实例
        mqProducer.shutdown();
    }
}

启动生产者,如下可看到,10条消息成功发送到Broker中。

SendResult [sendStatus=SEND_OK, msgId=AC6E004E14E418B4AAC28D8648020000, offsetMsgId=0A005A5600002A9F000000000000548C, messageQueue=MessageQueue [topic=FilterMessageTopic, brokerName=broker-a, queueId=1], queueOffset=3]
SendResult [sendStatus=SEND_OK, msgId=AC6E004E14E418B4AAC28D8648110001, offsetMsgId=0A005A5600002A9F000000000000555D, messageQueue=MessageQueue [topic=FilterMessageTopic, brokerName=broker-a, queueId=2], queueOffset=3]
SendResult [sendStatus=SEND_OK, msgId=AC6E004E14E418B4AAC28D86481D0002, offsetMsgId=0A005A5600002A9F000000000000562E, messageQueue=MessageQueue [topic=FilterMessageTopic, brokerName=broker-a, queueId=3], queueOffset=3]
SendResult [sendStatus=SEND_OK, msgId=AC6E004E14E418B4AAC28D8648220003, offsetMsgId=0A005A5600002A9F00000000000056FF, messageQueue=MessageQueue [topic=FilterMessageTopic, brokerName=broker-a, queueId=0], queueOffset=2]
SendResult [sendStatus=SEND_OK, msgId=AC6E004E14E418B4AAC28D8648290004, offsetMsgId=0A005A5600002A9F00000000000057D0, messageQueue=MessageQueue [topic=FilterMessageTopic, brokerName=broker-a, queueId=1], queueOffset=4]
SendResult [sendStatus=SEND_OK, msgId=AC6E004E14E418B4AAC28D8648300005, offsetMsgId=0A005A5600002A9F00000000000058A1, messageQueue=MessageQueue [topic=FilterMessageTopic, brokerName=broker-a, queueId=2], queueOffset=4]
SendResult [sendStatus=SEND_OK, msgId=AC6E004E14E418B4AAC28D8648370006, offsetMsgId=0A005A5600002A9F0000000000005972, messageQueue=MessageQueue [topic=FilterMessageTopic, brokerName=broker-a, queueId=3], queueOffset=4]
SendResult [sendStatus=SEND_OK, msgId=AC6E004E14E418B4AAC28D86483D0007, offsetMsgId=0A005A5600002A9F0000000000005A43, messageQueue=MessageQueue [topic=FilterMessageTopic, brokerName=broker-a, queueId=0], queueOffset=3]
SendResult [sendStatus=SEND_OK, msgId=AC6E004E14E418B4AAC28D8648430008, offsetMsgId=0A005A5600002A9F0000000000005B14, messageQueue=MessageQueue [topic=FilterMessageTopic, brokerName=broker-a, queueId=1], queueOffset=5]
SendResult [sendStatus=SEND_OK, msgId=AC6E004E14E418B4AAC28D8648490009, offsetMsgId=0A005A5600002A9F0000000000005BE5, messageQueue=MessageQueue [topic=FilterMessageTopic, brokerName=broker-a, queueId=2], queueOffset=5]

(2)、消费者订阅消息

主要是通过mqPushConsumer.subscribe("FilterMessageTopic", "TagA || TagC || TagD") 指定需要订阅的Tag,如果订阅所有Tag的话,则传入*即可。

public class MQConsumer {
    public static void main(String[] args) throws MQClientException {
        // 创建DefaultMQPushConsumer类并设定消费者名称
        DefaultMQPushConsumer mqPushConsumer = new DefaultMQPushConsumer("consumer-group-test");
        // 设置NameServer地址,如果是集群的话,使用分号;分隔开
        mqPushConsumer.setNamesrvAddr("10.0.90.86:9876");
        // 设置Consumer第一次启动是从队列头部开开发者_JAVA开发始消费还是队列尾部开始消费
        // 如果不是第一次启动,那么按照上次消费的位置继续消费
        mqPushConsumer.setConsumeFromWhere(ConsumeFromWhere.CONSUME_FROM_LAST_OFFSET);
        // 设置消费模型,集群还是广播,默认为集群
        mqPushConsumer.setMessageModel(MessageModel.CLUSTERING);
        // 消费者最小线程量
        mqPushConsumer.setConsumeThreadMin(5);
        // 消费者最大线程量
        mqPushConsumer.setConsumeThreadMax(10);
        // 设置一次消费消息的条数,默认是1
        mqPushConsumer.setConsumeMessageBATchMaxSize(1);
        // 订阅一个或者多个Topic,以及Tag来过滤需要消费的消息,如果订阅该主题下的所有tag,则使用*
        // 本例中,只订阅Tag为: TagA 、 TagC 、 TagD的消息
        mqPushConsumer.subscribe("FilterMessageTopic", "TagA || TagC || TagD");
        // 注册回调实现类来处理从broker拉取回来的消息
        mqPushConsumer.registerMessageListener(new MessageListenerConhttp://www.devze.comcurrently() {
            // 监听类实现MessageListenerConcurrently接口即可,重写consumeMessage方法接收数据
            @Override
            public ConsumeConcurrentlyStatus consumeMessage(List<MessageExt> msgList, ConsumeConcurrentlyContext consumeConcurrentlyContext) {
                MessageExt messageExt = msgList.get(0);
                String body = new String(messageExt.getBody(), StandardCharsets.UTF_8);
                System.out.println("消费者接收到消息: " + messageExt.toString() + "---消息内容为:" + body);
                // 标记该消息已经被成功消费
                return ConsumeConcurrentlyStatus.CONSUME_SUCCESS;
            }
        });
        // 启动消费者实例
        mqPushConsumer.start();
    }
}

如下,可看到消费者端接收到6条消息。

消费者接收到消息: MessageExt [brokerName=broker-a, queueId=0, storeSize=209, queueOffset=2, sysFlag=0, bornTimestamp=1646019187746, bornHost=/10.0.90.115:55652, storeTimestamp=1646019187082, storeHost=/10.0.90.86:10911, msgId=0A005A5600002A9F00000000000056FF, commitLogOffset=22271, bodyCRC=1188153005, reconsumeTimes=0, preparedTransactionOffset=0, toString()=Message{topic='FilterMessageTopic', flag=0, properties={MIN_OFFSET=0, MAX_OFFSET=4, CONSUME_START_TIME=1646019218458, UNIQ_KEY=AC6E004E14E418B4AAC28D8648220003, CLUSTER=DefaultCluster, TAGS=TagD}, body=[104, 101, 108, 108, 111, 44, 32, -24, -65, -103, -26, -104, -81, -25, -84, -84, 52, -26, -99, -95, -26, -74, -120, -26, -127, -81], transactionId='null'}]---消息内容为:hello, 这是第4条消息

消费者接收到消息: MessageExt [brokerName=broker-a, queueId=0, storeSize=209, queueOffset=3, sysFlag=0, bornTimestamp=1646019187773, bornHost=/10.0.90.115:55652, storeTimestamp=1646019187109, storeHost=/10.0.90.86:10911, msgId=0A005A5600002A9F0000000000005A43, commitLogOffset=23107, bodyCRC=1559045667, reconsumeTimes=0, preparedTransactionOffset=0, toString()=Message{topic='FilterMessageTopic', flag=0, properties={MIN_OFFSET=0, MAX_OFFSET=4, CONSUME_START_TIME=1646019218458, UNIQ_KEY=AC6E004E14E418B4AAC28D86483D0007, CLUSTER=DefaultCluster, TAGS=TagC}, body=[104, 101, 108, 108, 111, 44, 32, -24, -65, -103, -26, -104, -81, -25, -84, -84, 56, -26, -99, -95, -26, -74, -120, -26, -127, -81], transactionId='null'}]---消息内容为:hello, 这是第8条消息

消费者接收到消息: MessageExt [brokerName=broker-a, queueId=1, storeSize=209, queueOffset=5, sysFlag=0, bornTimestamp=1646019187779, bornHost=/10.0.90.115:55652, storeTimestamp=1646019187115, storeHost=/10.0.90.86:10911, msgId=0A005A5600002A9F0000000000005B14, commitLogOffset=23316, bodyCRC=858737949, reconsumeTimes=0, preparedTransactionOffset=0, toString()=Message{topic='FilterMessageTopic', flag=0, properties={MIN_OFFSET=0, MAX_OFFSET=6, CONSUME_START_TIME=1646019219467, UNIQ_KEY=AC6E004E14E418B4AAC28D8648430008, CLUSTER=DefaultCluster, TAGS=TagD}, body=[104, 101, 108, 108, 111, 44, 32, -24, -65, -103, -26, -104, -81, -25, -84, -84, 57, -26, -99, -95, -26, -74, -120, -26, -127, -81], transactionId='null'}]---消息内容为:hello, 这是第9条消息

消费者接收到消息: MessageExt [brokerName=broker-a, queueId=1, storeSize=209, queueOffset=3, sysFlag=0, bornTimestamp=1646019187715, bornHost=/10.0.90.115:55652, storeTimestamp=1646019187057, storeHost=/10.0.90.86:10911, msgId=0A005A5600002A9F000000000000548C, commitLogOffset=21644, bodyCRC=553127401, reconsumeTimes=0, preparedTransactionOffset=0, toString()=Message{topic='FilterMessageTopic', flag=0, properties={MIN_OFFSET=0, MAX_OFFSET=6, CONSUME_START_TIME=1646019219468, UNIQ_KEY=AC6E004E14E418B4AAC28D8648020000, CLUSTER=DefaultCluster, TAGS=TagA}, body=[104, 101, 108, 108, 111, 44, 32, -24, -65, -103, -26, -104, -81, -25, -84, -84, 49, -26, -99, -95, -26, -74, -120, -26, -127, -81], transactionId='null'}]---消息内容为:hello, 这是第1条消息

消费者接收到消息: MessageExt [brokerName=broker-a, queueId=3, storeSize=209, queueOffset=3, sysFlag=0, bornTimestamp=1646019187741, bornHost=/10.0.90.115:55652, storeTimestamp=1646019187077, storeHost=/10.0.90.86:10911, msgId=0A005A5600002A9F000000000000562E, commitLogOffset=22062, bodyCRC=604888532, reconsumeTimes=0, preparedTransactionOffset=0, toString()=Message{topic='FilterMessageTopic', flag=0, properties={MIN_OFFSET=0, MAX_OFFSET=5, CONSUME_START_TIME=1646019219472, UNIQ_KEY=AC6E004E14E418B4AAC28D86481D0002, CLUSTER=DefaultCluster, TAGS=TagC}, body=[104, 101, 108, 108, 111, 44, 32, -24, -65, -103, -26, -104, -81, -25, -84, -84, 51python, -26, -99, -95, -26, -74, -120, -26, -127, -81], transactionId='null'}]---消息内容为:hello, 这是第3条消息

消费者接收到消息: MessageExt [brokerName=broker-a, queueId=2, storeSize=209, queueOffset=4, sysFlag=0, bornTimestamp=1646019187760, bornHost=/10.0.90.115:55652, storeTimestamp=1646019187097, storeHost=/10.0.90.86:10911, msgId=0A005A5600002A9F00000000000058A1, commitLogOffset=22689, bodyCRC=1109661328, reconsumeTimes=0, preparedTransactionOffset=0, toString()=Message{topic='FilterMessageTopic', flag=0, properties={MIN_OFFSET=0, MAX_OFFSET=6, CONSUME_START_TIME=1646019219473, UNIQ_KEY=AC6E004E14E418B4AAC28D8648300005, CLUSTER=DefaultCluster, TAGS=TagA}, body=[104, 101, 108, 108, 111, 44, 32, -24, -65, -103, -26, -104, -81, -25, -84, -84, 54, -26, -99, -95, -26, -74, -120, -26, -127, -81], transactionId='null'}]---消息内容为:hello, 这是第6条消息

具体分析如下:

// 消息发送时总共5个Tag
String[] tags = new String[]{"TagA", "TagB", "TagC", "TagD", "TagE"};
// Tag计算方法
tag = i % tags.length = i % 5
  • 第1条消息,i = 0,消息的标签tag = tags[i % tags.length] = tags[0] = TagA;
  • 第2条消息,i = 1,消息的标签tag = tags[i % tags.length] = tags[1] = TagB;
  • 第3条消息,i = 2,消息的标签tag = tags[i % tags.length] = tags[2] = TagC;
  • 第4条消息,i = 3,消息的标签tag = tags[i % tags.length] = tags[3] = TagD;
  • 第5条消息,i = 4,消息的标签tag = tags[i % tags.length] = tags[4] = TagE;
  • 第6条消息,i = 5,消息的标签tag = tags[i % tags.length] = tags[0] = TagA;
  • 第7条消息,i = 6,消息的标签tag = tags[i % tags.length] = tags[1] = TagB;
  • 第8条消息,i = 7,消息的标签tag = tags[i % tags.length] = tags[2] = TagC;
  • 第9条消息,i = 8,消息的标签tag = tags[i % tags.length] = tags[3] = TagD;
  • 第10条消息,i = 9,消息的标签tag = tags[i % tags.length] = tags[4] = TagE;

因为消费者端只订阅了 TagA 、 TagC 、 TagD的消息,所以对应上面的,消费者端只会收到六条消息,即第1、3、4、6、8、9条消息。

三、根据自定义属性进行过滤 (SQL过滤)

通过Tag过滤消息可以很方便地选择您想要的消息,但是对于比较复杂的场合,使用Tag过滤的话可能不太满足条件。在这种情况下,可以使用SQL表达式筛选消息。SQL特性可以通过发送消息时的属性来进行计算。

RocketMQ只定义了一些基本语法来支持这个特性。

  • 数值比较,比如:>,>=,<,<=,BETWEEN,=;
  • 字符比较,比如:=,<>,IN;
  • IS NULL 或者 IS NOT NULL;
  • 逻辑符号 AND,OR,NOT;

常量支持类型为:

  • 数值,比如:123,3.1415;
  • 字符,比如:'abc',必须用单引号包裹起来;
  • NULL,特殊的常量
  • 布尔值,TRUE 或 FALSE

注意,只有使用push推送模式的消费者才能用使用SQL92标准的sql语句,pull拉取模式的消费者是不支持这个功能的。

下面我们通过一个示例演示:

(1)、生产者发送消息

生产者发送消息时,通过putUserProperty来设置消息的属性,实际上就是通过一个Map将用户自定义的属性保存到消息的properties属性中。

public class MQProducer {
    public static void main(String[] args) throws MQClientException, UnsupportedEncodingException, RemotingException, InterruptedException, MQBrokerException {
        // 创建DefaultMQProducer类并设定生产者名称
        DefaultMQProducer mqProducer = new DefaultMQProducer("producer-group-test");
        // 设置NameServer地址,如果是集群的话,使用分号;分隔开
        mqProducer.setNamesrvAddr("10.0.90.86:9876");
        // 消息最大长度 默认4M
        mqProducer.setMaxMessageSize(4096);
        // 发送消息超时时间,默认3000
        mqProducer.setSendMsgTimeout(3000);
        // 发送消息失败重试次数,默认2
        mqProducer.setRetryTimesWhenSendAsyncFailed(2);
        // 启动消息生产者
        mqProducer.start();
        for (int i = 0; i < 10; i++) {
            String msg = "hello, 这是第" + (i + 1) + "条消息";
            // 创建消息,并指定Topic(主题),Tag(标签)和消息内容
            Message message = new Message("FilterMessageTopic", "", msg.getBytes(RemotingHelper.DEFAULT_CHARSET));
            // 设置用户的一些自定义属性,本质上就是保存到一个Map中:private Map<String, String> properties
            message.putUserProperty("num", String.valueOf(i));
            message.putUserProperty("info", i % 2 == 0 ? "aaa" : "bbb");
            // 发送同步消息到一个Broker,可以通过sendResult返回消息是否成功送达
            SendResult sendResult = mqProducer.send(message);
            System.out.println(sendResult);
        }
        // 如果不再发送消息,关闭Producer实例
        mqProducer.shutdown();
    }
}

SendResult [sendStatus=SEND_OK, msgId=AC6E00924AE418B4AAC28DE709D90000, offsetMsgId=0A005A5600002A9F0000000000006E4E, messageQueue=MessageQueue [topic=FilterMessageTopic, brokerName=broker-a, queueId=2], queueOffset=11]

SendResult [sendStatus=SEND_OK, msgId=AC6E00924AE418B4AAC28DE709EF0001, offsetMsgId=0A005A5600002A9F0000000000006F24, messageQueue=MessageQueue [topic=FilterMessageTopic, brokerName=broker-a, queueId=3], queueOffset=11]

SendResult [sendStatus=SEND_OK, msgId=AC6E00924AE418B4AAC28DE709F50002, of编程客栈fsetMsgId=0A005A5600002A9F0000000000006FFA, messageQueue=MessageQueue [topic=FilterMessageTopic, brokerName=broker-a, queueId=0], queueOffset=9]

SendResult [sendStatus=SEND_OK, msgId=AC6E00924AE418B4AAC28DE709FB0003, offsetMsgId=0A005A5600002A9F00000000000070D0, messageQueue=MessageQueue [topic=FilterMessageTopic, brokerName=broker-a, queueId=1], queueOffset=11]

SendResult [sendStatus=SEND_OK, msgId=AC6E00924AE418B4AAC28DE70A020004, offsetMsgId=0A005A5600002A9F00000000000071A6, messageQueue=MessageQueue [topic=FilterMessageTopic, brokerName=broker-a, queueId=2], queueOffset=12]

SendResult [sendStatus=SEND_OK, msgId=AC6E00924AE418B4AAC28DE70A0C0005, offsetMsgId=0A005A5600002A9F000000000000727C, messageQueue=MessageQueue [topic=FilterMessageTopic, brokerName=broker-a, queueId=3], queueOffset=12]

SendResult [sendStatus=SEND_OK, msgId=AC6E00924AE418B4AAC28DE70A120006, offsetMsgId=0A005A5600002A9F0000000000007352, messageQueue=MessageQueue [topic=FilterMessageTopic, brokerName=broker-a, queueId=0], queueOffset=10]

SendResult [sendStatus=SEND_OK, msgId=AC6E00924AE418B4AAC28DE70A1C0007, offsetMsgId=0A005A5600002A9F0000000000007428, messageQueue=MessageQueue [topic=FilterMessageTopic, brokerName=broker-a, queueId=1], queueOffset=12]

SendResult [sendStatus=SEND_OK, msgId=AC6E00924AE418B4AAC28DE70A210008, offsetMsgId=0A005A5600002A9F00000000000074FE, messageQueue=MessageQueue [topic=FilterMessageTopic, brokerName=broker-a, queueId=2], queueOffset=13]

SendResult [sendStatus=SEND_OK, msgId=AC6E00924AE418B4AAC28DE70A270009, offsetMsgId=0A005A5600002A9F00000000000075D4, messageQueue=MessageQueue [topic=FilterMessageTopic, brokerName=broker-a, queueId=3], queueOffset=13]

(2)、消费者消费消息

消费者端使用如下接口指定SQL过滤的语法:

public void subscribe(finalString topic, final MessageSelector messageSelector)
// 用MessageSelector.bySql来使用sql筛选消息 
MessageSelector messageSelector = MessageSelector.bySql("xxxx");
public class MQConsumer {
    public static void main(String[] args) throws MQClientException {
        // 创建DefaultMQPushConsumer类并设定消费者名称
        DefaultMQPushConsumer mqPushConsumer = new DefaultMQPushConsumer("consumer-group-test");
        // 设置NameServer地址,如果是集群的话,使用分号;分隔开
        mqPushConsumer.setNamesrvAddr("10.0.90.86:9876");
        // 设置Consumer第一次启动是从队列头部开始消费还是队列尾部开始消费
        // 如果不是第一次启动,那么按照上次消费的位置继续消费
        mqPushConsumer.setConsumeFromWhere(ConsumeFromWhere.CONSUME_FROM_LAST_OFFSET);
        // 设置消费模型,集群还是广播,默认为集群
        mqPushConsumer.setMessageModel(MessageModel.CLUSTERING);
        // 消费者最小线程量
        mqPushConsumer.setConsumeThreadMin(5);
        // 消费者最大线程量
        mqPushConsumer.setConsumeThreadMax(10);
        // 设置一次消费消息的条数,默认是1
        mqPushConsumer.setConsumeMessageBatchMaxSize(1);
        // 用MessageSelector.bySql来使用sql筛选消息
        mqPushConsumer.subscribe("FJrkWaHJvzcilterMessageTopic", MessageSelector.bySql("(num between 0 and 5 ) and (info = 'aaa')"));
        // 注册回调实现类来处理从broker拉取回来的消息
        mqPushConsumer.registerMepythonssageListener(new MessageListenerConcurrently() {
            // 监听类实现MessageListenerConcurrently接口即可,重写consumeMessage方法接收数据
            @Override
            public ConsumeConcurrentlyStatus consumeMessage(List<MessageExt> msgList, ConsumeConcurrentlyContext consumeConcurrentlyContext) {
                MessageExt messageExt = msgList.get(0);
                String body = new String(messageExt.getBody(), StandardCharsets.UTF_8);
                System.out.println("消费者接收到消息: " + messageExt.toString() + "---消息内容为:" + body);
                // 标记该消息已经被成功消费
                return ConsumeConcurrentlyStatus.CONSUME_SUCCESS;
            }
        });
        // 启动消费者实例
        mqPushConsumer.start();
    }
}

我们直接运行消费者,发现启动报错了,如下:

Exception in thread "main" org.apache.rocketmq.client.exception.MQClientException: CODE: 1  DESC: The broker does not support consumer to filter message by SQL92

这个错误是由于RocketMQ默认是关闭了属性过滤功能的,如果需要使用该功能,需要开启enablePropertyFilter的属性,将该属性置为true才可以。也就是我们需要在RocketMQ的配置文件中添加如下配置:

// 开启属性过滤功能

enablePropertyFilter=true

重新启动RocketMQ后,再次运行消费者,如下可看到,消费者接收到三条消息:

消费者接收到消息: MessageExt [brokerName=broker-a, queueId=2, storeSize=214, queueOffset=11, sysFlag=0, bornTimestamp=1646025528795, bornHost=/10.0.90.115:59083, storeTimestamp=1646025527395, storeHost=/10.0.90.86:10911, msgId=0A005A5600002A9F0000000000006E4E, commitLogOffset=28238, bodyCRC=553127401, reconsumeTimes=0, preparedTransactionOffset=0, toString()=Message{topic='FilterMessageTopic', flag=0, properties={MIN_OFFSET=0, MAX_OFFSET=14, num=0, CONSUME_START_TIME=1646025559201, UNIQ_KEY=AC6E00924AE418B4AAC28DE709D90000, CLUSTER=DefaultCluster, info=aaa}, body=[104, 101, 108, 108, 111, 44, 32, -24, -65, -103, -26, -104, -81, -25, -84, -84, 49, -26, -99, -95, -26, -74, -120, -26, -127, -81], transactionId='null'}]---消息内容为:hello, 这是第1条消息

消费者接收到消息: MessageExt [brokerName=broker-a, queueId=0, storeSize=214, queueOffset=9, sysFlag=0, bornTimestamp=1646025528821, bornHost=/10.0.90.115:59083, storeTimestamp=1646025527417, storeHost=/10.0.90.86:10911, msgId=0A005A5600002A9F0000000000006FFA, commitLogOffset=28666, bodyCRC=604888532, reconsumeTimes=0, preparedTransactionOffset=0, toString()=Message{topic='FilterMessageTopic', flag=0, properties={MIN_OFFSET=0, MAX_OFFSET=11, num=2, CONSUME_START_TIME=1646025559201, UNIQ_KEY=AC6E00924AE418B4AAC28DE709F50002, CLUSTER=DefaultCluster, info=aaa}, body=[104, 101, 108, 108, 111, 44, 32, -24, -65, -103, -26, -104, -81, -25, -84, -84, 51, -26, -99, -95, -26, -74, -120, -26, -127, -81], transactionId='null'}]---消息内容为:hello, 这是第3条消息

消费者接收到消息: MessageExt [brokerName=broker-a, queueId=2, storeSize=214, queueOffset=12, sysFlag=0, bornTimestamp=1646025528834, bornHost=/10.0.90.115:59083, storeTimestamp=1646025527433, storeHost=/10.0.90.86:10911, msgId=0A005A5600002A9F00000000000071A6, commitLogOffset=29094, bodyCRC=689155475, reconsumeTimes=0, preparedTransactionOffset=0, toString()=Message{topic='FilterMessageTopic', flag=0, properties={MIN_OFFSET=0, MAX_OFFSET=14, num=4, CONSUME_START_TIME=1646025559201, UNIQ_KEY=AC6E00924AE418B4AAC28DE70A020004, CLUSTER=DefaultCluster, info=aaa}, body=[104, 101, 108, 108, 111, 44, 32, -24, -65, -103, -26, -104, -81, -25, -84, -84, 53, -26, -99, -95, -26, -74, -120, -26, -127, -81], transactionId='null'}]---消息内容为:hello, 这是第5条消息

分析:

生产者发送消息的时候,添加了用户自定义属性num、info,通过上述控制台输出消息的properties属性我们也可以看到。num的值其实就是0-9,info的值是偶数的时候为aaa,奇数的时候为bbb。消费者通过MessageSelector.bySql("(num between 0 and 5 ) and (info = 'aaa')")指定的过滤条件是:num在[0,5]之间并且info的值为aaa。因此,同时满足这两个条件的就只有三条消息。

到此这篇关于RocketMQ根据Tag进行消息过滤的文章就介绍到这了,更多相关RocketMQ消息过滤内容请搜索我们以前的文章或继续浏览下面的相关文章希望大家以后多多支持我们!

0

精彩评论

暂无评论...
验证码 换一张
取 消