环境:springboot2.3.12.RELEASE + kafka_2.13-2.7.0 + zookeeper-3.6.2
Kafka Stream介绍
Kafka在0.10版本推出了Stream API,提供了对存储在Kafka内的数据进行流式处理和分析的能力。
流式计算一般被用来和批量计算做比较。批量计算往往有一个固定的数据集作为输入并计算结果。而流式计算的输入往往是“无界”的(Unbounded Data),持续输入的,即永远拿不到全量数据去做计算;同时,计算结果也是持续输出的,只能拿到某一个时刻的结果,而不是最终的结果。
Kafka Streams是一个客户端类库,用于处理和分析存储在Kafka中的数据。它建立在流式处理的一些重要的概念之上:如何区分事件时间和处理时间、Windowing的支持、简单高效的管理和实时查询应用程序状态。
Kafka Streams的门槛非常低:和编写一个普通的Kafka消息处理程序没有太大的差异,可以通过多进程部署来完成扩容、负载均衡、高可用(Kafka Consumer的并行模型)。
Kafka Streams的一些特点:
- 被设计成一个简单的、轻量级的客户端类库,能够被集成到任何Java应用中
- 除了Kafka之外没有任何额外的依赖,利用Kafka的分区模型支持水平扩容和保证顺序性
- 通过可容错的状态存储实现高效的状态操作(windowed joins and aggregations)
- 支持exactly-once语义
- 支持纪录级的处理,实现毫秒级的延迟
- 提供High-Level的Stream DSL和Low-Level的Processor API
Stream Processing Topology流处理拓扑
- 流是Kafka Streams提供的最重要的抽象:它表示一个无限的、不断更新的数据集。流是不可变数据记录的有序、可重放和容错序列,其中数据记录定义为键值对。
- Stream Processing Application是使用了Kafka Streams库的应用程序。它通过processor topologies定义计算逻辑,其中每个processor topology都是多个stream processor(节点)通过stream组成的图。
- A stream processor 是处理器拓扑中的节点;它表示一个处理步骤,通过每次从拓扑中的上游处理器接收一个输入记录,将其操作应用于该记录,来转换流中的数据,并且随后可以向其下游处理器生成一个或多个输出记录。
有两种特殊的processor:
Source Processor 源处理器是一种特殊类型的流处理器,它没有任何上游处理器。它通过使用来自一个或多个kafka topic的记录并将其转发到其下游处理器,从而从一个或多个kafka topic生成其拓扑的输入流。
Sink Processor 接收器处理器是一种特殊类型的流处理器,没有下游处理器。它将从其上游处理器接收到的任何记录发送到指定的kafka topic。
相关的核心概念查看如下链接
下面演示Kafka Stream 在Springboot中的应用
依赖
- <dependency>
- <groupId>org.springframework.boot</groupId>
- <artifactId>spring-boot-starter-web</artifactId>
- </dependency>
- <dependency>
- <groupId>org.springframework.kafka</groupId>
- <artifactId>spring-kafka</artifactId>
- </dependency>
- <dependency>
- <groupId>org.apache.kafka</groupId>
- <artifactId>kafka-streams</artifactId>
- </dependency>
配置
- server:
- port: 9090
- spring:
- application:
- name: kafka-demo
- kafka:
- streams:
- application-id: ${spring.application.name}
- properties:
- spring.json.trusted.packages: '*'
- bootstrap-servers:
- - localhost:9092
- - localhost:9093
- - localhost:9094
- producer:
- acks: 1
- retries: 10
- key-serializer: org.apache.kafka.common.serialization.StringSerializer
- value-serializer: org.springframework.kafka.support.serializer.JsonSerializer #org.apache.kafka.common.serialization.StringSerializer
- properties:
- spring.json.trusted.packages: '*'
- consumer:
- key-deserializer: org.apache.kafka.common.serialization.StringDeserializer
- value-deserializer: org.springframework.kafka.support.serializer.JsonDeserializer #org.apache.kafka.common.serialization.StringDeserializer
- enable-auto-commit: false
- group-id: ConsumerTest
- auto-offset-reset: latest
- properties:
- session.timeout.ms: 12000
- heartbeat.interval.ms: 3000
- max.poll.records: 100
- spring.json.trusted.packages: '*'
- listener:
- ack-mode: manual-immediate
- type: batch
- concurrency: 8
- properties:
- max.poll.interval.ms: 300000
消息发送
- @Service
- public class MessageSend {
- @Resource
- private KafkaTemplate<String, Message> kafkaTemplate ;
- public void sendMessage2(Message message) {
- kafkaTemplate.send(new ProducerRecord<String, Message>("test", message)).addCallback(result -> {
- System.out.println("执行成功..." + Thread.currentThread().getName()) ;
- }, ex -> {
- System.out.println("执行失败") ;
- ex.printStackTrace() ;
- }) ;
- }
- }
消息监听
- @KafkaListener(topics = {"test"})
- public void listener2(List<ConsumerRecord<String, Message>> records, Acknowledgment ack) {
- for (ConsumerRecord<String, Message> record : records) {
- System.out.println(this.getClass().hashCode() + ", Thread" + Thread.currentThread().getName() + ", key: " + record.key() + ", 接收到消息:" + record.value() + ", patition: " + record.partition() + ", offset: " + record.offset()) ;
- }
- try {
- TimeUnit.SECONDS.sleep(0) ;
- } catch (InterruptedException e) {
- e.printStackTrace();
- }
- ack.acknowledge() ;
- }
- @KafkaListener(topics = {"demo"})
- public void listenerDemo(List<ConsumerRecord<String, Message>> records, Acknowledgment ack) {
- for (ConsumerRecord<String, Message> record : records) {
- System.out.println("Demo Topic: " + this.getClass().hashCode() + ", Thread" + Thread.currentThread().getName() + ", key: " + record.key() + ", 接收到消息:" + record.value() + ", patition: " + record.partition() + ", offset: " + record.offset()) ;
- }
- ack.acknowledge() ;
- }
Kafka Stream处理
消息转换并转发其它Topic
- @Bean
- public KStream<Object, Object> kStream(StreamsBuilder streamsBuilder) {
- KStream<Object, Object> stream = streamsBuilder.stream("test");
- stream.map((key, value) -> {
- System.out.println("原始消息内容:" + new String((byte[]) value, Charset.forName("UTF-8"))) ;
- return new KeyValue<>(key, "{\"title\": \"123123\", \"message\": \"重新定义内容\"}".getBytes(Charset.forName("UTF-8"))) ;
- }).to("demo") ;
- return stream;
- }
执行结果:
Stream对象处理
- @Bean
- public KStream<String, Message> kStream4(StreamsBuilder streamsBuilder) {
- JsonSerde<Message> jsonSerde = new JsonSerde<>() ;
- JsonDeserializer<Message> descri = (JsonDeserializer<Message>) jsonSerde.deserializer() ;
- descri.addTrustedPackages("*") ;
- KStream<String, Message> stream = streamsBuilder.stream("test", Consumed.with(Serdes.String(), jsonSerde));
- stream.map((key, value) -> {
- value.setTitle("XXXXXXX") ;
- return new KeyValue<>(key, value) ;
- }).to("demo", Produced.with(Serdes.String(), jsonSerde)) ;
- return stream;
- }
执行结果:
分组处理
- @Bean
- public KStream<String, Message> kStream5(StreamsBuilder streamsBuilder) {
- JsonSerde<Message> jsonSerde = new JsonSerde<>() ;
- JsonDeserializer<Message> descri = (JsonDeserializer<Message>) jsonSerde.deserializer() ;
- descri.addTrustedPackages("*") ;
- KStream<String, Message> stream = streamsBuilder.stream("test", Consumed.with(Serdes.String(), jsonSerde));
- stream.selectKey(new KeyValueMapper<String, Message, String>() {
- @Override
- public String apply(String key, Message value) {
- return value.getOrgCode() ;
- }
- })
- .groupByKey(Grouped.with(Serdes.String(), jsonSerde))
- .count()
- .toStream().print(Printed.toSysOut());
- return stream;
- }
执行结果:
聚合
- @Bean
- public KStream<String, Message> kStream6(StreamsBuilder streamsBuilder) {
- JsonSerde<Message> jsonSerde = new JsonSerde<>() ;
- JsonDeserializer<Message> descri = (JsonDeserializer<Message>) jsonSerde.deserializer() ;
- descri.addTrustedPackages("*") ;
- KStream<String, Message> stream = streamsBuilder.stream("test", Consumed.with(Serdes.String(), jsonSerde));
- stream.selectKey(new KeyValueMapper<String, Message, String>() {
- @Override
- public String apply(String key, Message value) {
- return value.getOrgCode() ;
- }
- })
- .groupByKey(Grouped.with(Serdes.String(), jsonSerde))
- .aggregate(() -> 0L, (key, value ,aggValue) -> {
- System.out.println("key = " + key + ", value = " + value + ", agg = " + aggValue) ;
- return aggValue + 1 ;
- }, Materialized.<String, Long, KeyValueStore<Bytes,byte[]>>as("kvs").withValueSerde(Serdes.Long()))
- .toStream().print(Printed.toSysOut());
- return stream;
- }
执行结果:
Filter过滤数据
- @Bean
- public KStream<String, Message> kStream7(StreamsBuilder streamsBuilder) {
- JsonSerde<Message> jsonSerde = new JsonSerde<>() ;
- JsonDeserializer<Message> descri = (JsonDeserializer<Message>) jsonSerde.deserializer() ;
- descri.addTrustedPackages("*") ;
- KStream<String, Message> stream = streamsBuilder.stream("test", Consumed.with(Serdes.String(), jsonSerde));
- stream.selectKey(new KeyValueMapper<String, Message, String>() {
- @Override
- public String apply(String key, Message value) {
- return value.getOrgCode() ;
- }
- })
- .groupByKey(Grouped.with(Serdes.String(), jsonSerde))
- .aggregate(() -> 0L, (key, value ,aggValue) -> {
- System.out.println("key = " + key + ", value = " + value + ", agg = " + aggValue) ;
- return aggValue + 1 ;
- }, Materialized.<String, Long, KeyValueStore<Bytes,byte[]>>as("kvs").withValueSerde(Serdes.Long()))
- .toStream()
- .filter((key, value) -> !"2".equals(key))
- .print(Printed.toSysOut());
- return stream;
- }
执行结果:
过滤Key不等于"2"
分支多流处理
- @Bean
- public KStream<String, Message> kStream8(StreamsBuilder streamsBuilder) {
- JsonSerde<Message> jsonSerde = new JsonSerde<>() ;
- JsonDeserializer<Message> descri = (JsonDeserializer<Message>) jsonSerde.deserializer() ;
- descri.addTrustedPackages("*") ;
- KStream<String, Message> stream = streamsBuilder.stream("test", Consumed.with(Serdes.String(), jsonSerde));
- // 分支,多流处理
- KStream<String, Message>[] arrStream = stream.branch(
- (key, value) -> "男".equals(value.getSex()),
- (key, value) -> "女".equals(value.getSex()));
- Stream.of(arrStream).forEach(as -> {
- as.foreach((key, message) -> {
- System.out.println(Thread.currentThread().getName() + ", key = " + key + ", message = " + message) ;
- });
- });
- return stream;
- }
执行结果:
多字段分组
不能使用多个selectKey,后面的会覆盖前面的
- @Bean
- public KStream<String, Message> kStreamM2(StreamsBuilder streamsBuilder) {
- JsonSerde<Message> jsonSerde = new JsonSerde<>() ;
- JsonDeserializer<Message> descri = (JsonDeserializer<Message>) jsonSerde.deserializer() ;
- descri.addTrustedPackages("*") ;
- KStream<String, Message> stream = streamsBuilder.stream("test", Consumed.with(Serdes.String(), jsonSerde));
- stream
- .selectKey(new KeyValueMapper<String, Message, String>() {
- @Override
- public String apply(String key, Message value) {
- System.out.println(Thread.currentThread().getName()) ;
- return value.getTime() + " | " + value.getOrgCode() ;
- }
- })
- .groupByKey(Grouped.with(Serdes.String(), jsonSerde))
- .count()
- .toStream().print(Printed.toSysOut());
- return stream;
- }
执行结果: