前言:
这是一个Flink自定义开发的基础教学。本文将通过flink的DataStream模块API,以kafka为数据源,构建一个基础测试环境;包含一个kafka生产者线程工具,一个自定义FilterFunction算子,一个自定义MapFunction算子,用一个flink任务的代码逻辑,将实时读kafka并多层处理串起来;让读者体会通过Flink构建自定义函数的技巧。
一、Flink的开发模块分析
Flink提供四个基础模块:核心SDK开发API分别是处理实时计算的DataStream和处理离线计算的DataSet;基于这两个SDK,在其上包装了TableAPI开发模块的SDK;在Table API之上,定义了高度抽象可用SQL开发任务的FlinkSQL。在核心开发API之下,还有基础API的接口,可用于对时间,状态,算子等最细粒度的特性对象做操作,如包装自定义算子的ProcessWindowFunction和ProcessFunction等基础函数以及内置的对象状态StateTtlConfig;
FLINK开发API关系结构如下:
二、定制化开发Demo演示
2.1 场景介绍
Flink实时任务的的通用技术架构是消息队列中间件+Flink任务:
将数据采集到Kafka或pulser这类队列中间件的Topic,然后使用Flink内置的kafkaSource,监控Topic的数据情况,做实时处理。
- 这里提供一个kafka的生产者线程,可以自定义构建需要的数据和上传时间,用于控制写入kafka的数据源;
- 重写两个DataStream的基础算子:FilterFunction和MapFunction,用于让读者体会,如何对FLINK函数的重新包装,后续更基础的函数原理一样;我这里用String数据对象做处理,减少对象转换的SDK引入,通常要基于业务做数据polo的加工,这个自己处理,将对象换成业务对象;
- 然后使用Flink将整个业务串起来,从kafka读数据,经过两层处理,最终输出需要的结果;
2.2 本地demo演示
2.2.1 pom文件
这里以flink1.14.6+scala1.12版本为例:
2.2.2 kafka生产者线程方法
package org.example.util;import org.apache.kafka.clients.producer.KafkaProducer;
import org.apache.kafka.clients.producer.Producer;
import org.apache.kafka.clients.producer.ProducerRecord;import java.util.*;/*** 向kafka生产数据** @author i7杨* @date 2024/01/12 13:02:29*/public class KafkaProducerUtil extends Thread {private String topic;public KafkaProducerUtil(String topic) {super();this.topic = topic;}private static Producer<String, String> createProducer() {// 通过Properties类设置Producer的属性Properties properties = new Properties();
// 测试环境 kafka 配置properties.put("bootstrap.servers", "ip2:9092,ip:9092,ip3:9092");properties.put("key.serializer", "org.apache.kafka.common.serialization.StringSerializer");properties.put("value.serializer", "org.apache.kafka.common.serialization.StringSerializer");return new KafkaProducer<String, String>(properties);}@Overridepublic void run() {Producer<String, String> producer = createProducer();Random random = new Random();Random random2 = new Random();while (true) {int nums = random.nextInt(10);int nums2 = random.nextInt(50);
// double nums2 = random2.nextDouble();String time = new Date().getTime() / 1000 + 5 + "";String type = "pv";try {if (nums2 % 2 == 0) {type = "pv";} else {type = "uv";}
// String info = "{\"user\":" + nums + ",\"item\":" + nums * 10 + ",\"category\":" + nums2 + ",\"pv\":" + nums2 * 5 + ",\"ts\":\"" + time + "\"}";String info = nums + "=" + nums2;System.out.println("message : " + info);producer.send(new ProducerRecord<String, String>(this.topic, info));} catch (Exception e) {e.printStackTrace();}System.out.println("=========数据已经写入==========");try {sleep(1000);} catch (InterruptedException e) {e.printStackTrace();}}}public static void main(String[] args) {new KafkaProducerUtil("test01").run();}public static void sendMessage(String topic, String message) {Producer<String, String> producer = createProducer();producer.send(new ProducerRecord<String, String>(topic, message));}}
2.2.3 自定义基础函数
这里自定义了filter和map两个算子函数,测试逻辑按照数据结构变化:
自定义FilterFunction函数算子:阈值小于40的过滤掉
package org.example.funtion;import org.apache.flink.api.common.functions.FilterFunction;/*** FilterFunction重构** @author i7杨* @date 2024/01/12 13:02:29*/public class InfoFilterFunction implements FilterFunction<String> {private double threshold;public InfoFilterFunction(double threshold) {this.threshold = threshold;}@Overridepublic boolean filter(String value) throws Exception {if (value.split("=").length == 2)// 阈值过滤return Double.valueOf(value.split("=")[1]) > threshold;else return false;}
}
自定义MapFunction函数:后缀为2的,添加上特殊信息
package org.example.funtion;import org.apache.flink.api.common.functions.MapFunction;public class ActionMapFunction implements MapFunction<String, String> {@Overridepublic String map(String value) throws Exception {System.out.println("value:" + value);if (value.endsWith("2"))return value.concat(":Special processing information");else return value;}
}
2.2.4 flink任务代码
任务逻辑:使用kafka工具产生数据,然后监控kafka的topic,讲几个函数串起来,输出结果;
package org.example.service;import org.apache.flink.api.common.serialization.SimpleStringSchema;
import org.apache.flink.streaming.api.datastream.DataStream;
import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
import org.apache.flink.streaming.connectors.kafka.FlinkKafkaConsumer;
import org.apache.kafka.clients.consumer.ConsumerConfig;
import org.apache.kafka.common.serialization.StringDeserializer;
import org.example.funtion.ActionMapFunction;
import org.example.funtion.InfoFilterFunction;import java.util.*;public class FlinkTestDemo {public static void main(String[] args) throws Exception {// 设置执行环境StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();// Kafka 配置Properties kafkaProps = new Properties();kafkaProps.setProperty(ConsumerConfig.BOOTSTRAP_SERVERS_CONFIG, "ip1:9092,ip2:9092,ip3:9092");kafkaProps.setProperty(ConsumerConfig.GROUP_ID_CONFIG, "flink-consumer-group");kafkaProps.setProperty(ConsumerConfig.KEY_DESERIALIZER_CLASS_CONFIG, StringDeserializer.class.getName());kafkaProps.setProperty(ConsumerConfig.VALUE_DESERIALIZER_CLASS_CONFIG, StringDeserializer.class.getName());kafkaProps.setProperty(ConsumerConfig.AUTO_OFFSET_RESET_CONFIG, "earliest");// 创建 Kafka 消费者FlinkKafkaConsumer<String> kafkaConsumer = new FlinkKafkaConsumer<>("test01",// Kafka 主题名称new SimpleStringSchema(),kafkaProps);// 从 Kafka 中读取数据流DataStream<String> kafkaStream = env.addSource(kafkaConsumer);env.disableOperatorChaining();kafkaStream.filter(new InfoFilterFunction(40)).map(new ActionMapFunction()).print("阈值大于40以上的message=");// 执行任务env.execute("This is a testing task");}}
运行结果: