消息过滤包括基于表达式过滤与基于类模式两种过滤模式。其中表达式过滤又分为TAG和SQL92模式,分别介绍各自的过滤机制,及代码示例内容,深入探消息过滤的原理。
1、TAG模式过滤
发送消息时我们会为每一条消息设置TAG标签,同一大类中的消息放在一个主题TOPIC下,但是如果进行分类我们则可以根据TAG进行分类,每一类消费者可能不是关系某个主题下的所有消息,我们就可以通过TAG进行过滤,订阅关注的某一类数据。
1.1、producer
public class Producer {
public static void main(String[] args) throws MQClientException, InterruptedException {
DefaultMQProducer producer = new DefaultMQProducer("producer_test");
producer.setNamesrvAddr("10.10.12.203:9876;10.10.12.204:9876");
producer.start();
String[] tags = {"TagA","TagB","TagC","TagD"};
for (int i = 0; i < 40; i++) {
try {
String tag = tags[i % tags.length];
//构建消息
Message msg = new Message("GumxTest" /* Topic */,
tag /* Tag */,
("RocketMQ消息测试,消息的TAG="+tag+" == " + i).getBytes(RemotingHelper.DEFAULT_CHARSET)
);
SendResult sendResult = producer.send(msg);
System.out.printf("%s%n", sendResult);
} catch (Exception e) {
e.printStackTrace();
Thread.sleep(1000);
}
}
producer.shutdown();
}
}
主题GumxTest,标签分别是"TagA","TagB","TagC","TagD"每个分别发送10条消息
1.2、consumer
public class Consumer {
public static void main(String[] args){
try {
DefaultMQPushConsumer consumer = new DefaultMQPushConsumer();
consumer.setConsumerGroup("consumer_tags");
consumer.setNamesrvAddr("10.10.12.203:9876;10.10.12.204:9876");
consumer.subscribe("GumxTest", "TagA || TagC");
consumer.registerMessageListener(new MessageListenerConcurrently(){
@Override
public ConsumeConcurrentlyStatus consumeMessage(List paramList,
ConsumeConcurrentlyContext paramConsumeConcurrentlyContext) {
try {
for(MessageExt msg : paramList){
String msgbody = new String(msg.getBody(), "utf-8");
SimpleDateFormat sd = new SimpleDateFormat("YYYY-MM-dd HH:mm:ss");
Date date = new Date(msg.getStoreTimestamp());
System.out.println("Consumer1=== 存入时间 : "+ sd.format(date) +" == MessageBody: "+ msgbody);//输出消息内容
}
} catch (Exception e) {
e.printStackTrace();
return ConsumeConcurrentlyStatus.RECONSUME_LATER; //稍后再试
}
return ConsumeConcurrentlyStatus.CONSUME_SUCCESS; //消费成功
}
});
consumer.start();
System.out.println("Consumer===启动成功!");
} catch (Exception e) {
// TODO Auto-generated catch block
e.printStackTrace();
}
}
}
查看结果:
消费者组订阅相同的主题不同的Tag时,如果订阅是多个Tag则通过“||” 分割
同一个消费者组订阅的主题,Tag必须相同
2、SQL表达式过滤
SQL92表达式消息过滤,是通过消息的属性运行SQL过滤表达式进行条件匹配,消息发送时需要设置用户的属性putUserProperty方法设置属性。
支持的语法:
- 数值比较, 如
>
,>=
,<
,<=
,BETWEEN
,=
; - 字符比较, 如
=
,<>
,IN
; -
IS NULL
orIS NOT NULL
; - 逻辑连接符
AND
,OR
,NOT
;
支持的类型:
- 数值型, 如123, 3.1415;
- 字符型, 如 ‘abc’, 必须用单引号;
-
NULL
, 特殊常数; - 布尔值,
TRUE
orFALSE
;
2.1、producer
public class Producer {
public static void main(String[] args) throws MQClientException, InterruptedException {
DefaultMQProducer producer = new DefaultMQProducer("producer_test");
producer.setNamesrvAddr("10.10.12.203:9876;10.10.12.204:9876");
producer.start();
String[] tags = {"TagA","TagB","TagC","TagD"};
for (int i = 0; i < 40; i++) {
try {
String tag = tags[i % tags.length];
//构建消息
Message msg = new Message("GumxTest" /* Topic */,
tag /* Tag */,
("RocketMQ消息测试,消息的TAG="+tag+ ", 属性 age = " + i + ", == " + i).getBytes(RemotingHelper.DEFAULT_CHARSET)
);
msg.putUserProperty("age", i+"");
SendResult sendResult = producer.send(msg);
System.out.printf("%s%n", sendResult);
} catch (Exception e) {
e.printStackTrace();
Thread.sleep(1000);
}
}
producer.shutdown();
}
}
2.2、consumer
public class Consumer {
public static void main(String[] args){
try {
DefaultMQPushConsumer consumer = new DefaultMQPushConsumer();
consumer.setConsumerGroup("consumer_sql");
consumer.setNamesrvAddr("10.10.12.203:9876;10.10.12.204:9876");
consumer.subscribe("GumxTest", MessageSelector.bySql("age between 0 and 8"));
consumer.registerMessageListener(new MessageListenerConcurrently(){
@Override
public ConsumeConcurrentlyStatus consumeMessage(List paramList,
ConsumeConcurrentlyContext paramConsumeConcurrentlyContext) {
try {
for(MessageExt msg : paramList){
String msgbody = new String(msg.getBody(), "utf-8");
SimpleDateFormat sd = new SimpleDateFormat("YYYY-MM-dd HH:mm:ss");
Date date = new Date(msg.getStoreTimestamp());
System.out.println("Consumer=== 存入时间 : "+ sd.format(date) +" == MessageBody: "+ msgbody);//输出消息内容
}
} catch (Exception e) {
e.printStackTrace();
return ConsumeConcurrentlyStatus.RECONSUME_LATER; //稍后再试
}
return ConsumeConcurrentlyStatus.CONSUME_SUCCESS; //消费成功
}
});
consumer.start();
System.out.println("Consumer===启动成功!");
} catch (Exception e) {
// TODO Auto-generated catch block
e.printStackTrace();
}
}
}
启动时:
启动消费者时发现启动失败,提示不支持SQL92过滤,网上也查了一些资料多说都是说版本太低,但是我使用的是RocketMQ4.2.0版本,已经是很高的版本了。
分析源码发现BrokerConfig的配置类中有一个属性 private boolean enablePropertyFilter = false;默认属性过滤没有开启,然而SQL92就是通过属性来过滤的。问题找到了,我们需要配置broker的属性在broker配置文件中添加enablePropertyFilter =true,需要依次关闭集群中的Broker、NameSrv服务,配置好后依次启动NameSrv、Broker服务
再次启动,启动成功,查看其结果:
3、类过滤模式(基于4.2.0版本)
RocketMQ通过定义消息过滤类的接口实现消息过滤
3.1、producer
public class Producer {
public static void main(String[] args) throws MQClientException, InterruptedException {
DefaultMQProducer producer = new DefaultMQProducer("ProducerGroupName");
producer.setNamesrvAddr("10.10.12.203:9876;10.10.12.204:9876");
producer.start();
try {
for (int i = 0; i < 6000000; i++) {
Message msg = new Message("TopicFilter",// topic
"TagA",// tag
("Hello MetaQ age = " + i ).getBytes());// body
msg.putUserProperty("age", String.valueOf(i));
SendResult sendResult = producer.send(msg);
System.out.println(sendResult);
Thread.sleep(1000);
}
}
catch (Exception e) {
e.printStackTrace();
}
producer.shutdown();
}
}
3.2、consumer
public class Consumer {
public static void main(String[] args) throws InterruptedException, MQClientException, IOException {
DefaultMQPushConsumer consumer = new DefaultMQPushConsumer("ConsumerGroupFilter");
consumer.setNamesrvAddr("10.10.12.203:9876;10.10.12.204:9876");
// 使用Java代码,在服务器做消息过滤
String filterCode = MixAll.file2String("D:\\WorkSoft\\workspace\\rocketmq-example\\src\\main\\java\\cn\\gumx\\rocketmq\\filter\\MessageFilterImpl.java");
consumer.subscribe("TopicFilter1", "cn.gumx.rocketmq.filter.MessageFilterImpl",filterCode);
consumer.registerMessageListener(new MessageListenerConcurrently() {
public ConsumeConcurrentlyStatus consumeMessage(List paramList,
ConsumeConcurrentlyContext context) {
try {
for(MessageExt msg : paramList){
String msgbody = new String(msg.getBody(), "utf-8");
System.out.println(Thread.currentThread().getName() + " Receive New Messages: " + msgbody);
}
} catch (UnsupportedEncodingException e) {
e.printStackTrace();
}
return ConsumeConcurrentlyStatus.CONSUME_SUCCESS;
}
});
consumer.start();
System.out.println("Consumer Started.");
}
}
自定义消息的过滤类
public class MessageFilterImpl implements MessageFilter {
public boolean match(MessageExt msg, FilterContext arg1) {
String property = msg.getUserProperty("age");
if (property != null) {
int age = Integer.parseInt(property);
if ((age % 3) == 0 && (age > 10)) {
return true;
}
}
return false;
}
}
consumer启动后并没有出现预期的结果,查了资料也没有相关介绍,只是和我们代码一样的处理逻辑,查看源码发现,需要启动filter组件mqfiltersrv服务。
./mqfiltersrv -n 10.10.12.203:9876;10.10.12.204:9876 &
我们部署的是双主,mqfiltersrv服务都需要开启
查看服务端结果:
使用类消息过滤模式,需要额外需要启动filter组件mqfiltersrv服务,否则消费不了,每个broker都需要启动一个,相当于加了一层过滤层。
filtersrv 出现了。减少了 Broker 的负担,又减少了 Consumer 接收无用的消息。当然缺点也是有的,多了一层 filtersrv 网络开销
MessageFilterImpl消息过滤实现类中的代码最好不要带有中文防止错误
注意:RocketMQ4.3.1开始删除与 mqfilter 服务器相关的脚本,4.3.2删除客户端关于mqfilter 客户端代码,后面版本不支持该功能。