flink01

1. flink简介

1.1 什么是flink

Apache Flink是一个分布式大数据处理引擎，可以对有限数据流（如离线数据）和无限流数据及逆行有状态计算（不太懂）。可以部署在各种集群环境，对各种大小的数据规模进行快速计算。

 1.2 flink的架构体系

具体见文档

2. flink的安装

• 修改flink-conf.yaml

jobmanager.rpc.address: feng05   // 注意，此处冒号后需要空一格，并且参数要顶行写（yaml文件格式的规定，否则会报错）
taskmanager.numberOfTaskSlots: 2

• 将配置好的Flink安装包拷?贝到其他节点

for i in {4..7}; do scp -r flink-1.10.1/ feng05:$PWD; done

• 启动集群（standalone模式）

bin/start-cluster.sh

• 查看Java进程（jps）

StandaloneSessionClusterEntrypoint (JobManager，即Master)
TaskManagerRunner (TaskManager，即Worker)

• 访问JobManager的web管理界面

feng05:8081

3. flink提交任务的两种方式

第一种：通过web页面提交

 第二种：通过命令行提交

./flink run -m feng05:8081 -p 4 -c cn._51doit.flink.day1.HelloFlink  /root/flink-in-action-1.0-SNAPSHOT.jar --hostname feng05 --port 8888

4.flink快速入门

4.0 创建flink工程

• java形式（window上）

mvn archetype:generate -DarchetypeGroupId=org.apache.flink -DarchetypeArtifactId=flink-quickstart-java -DarchetypeVersion=1.10.1 -DgroupId=cn._51doit.flink -DartifactId=flink-java -Dversion=1.0 -Dpackage=cn._51doit.flink -DinteractiveMode=false

• scala形式

同理

• 也可以直接在IDEA上创建相应的maven项目，导入pom文件（这里jar的版本不好弄，所以直接用上面的命令更方便）

4.1 wordCount案例

StreamWordCount（匿名内部类的形式）

package cn._51doit.flink.day01;

import org.apache.flink.api.common.functions.FlatMapFunction;
import org.apache.flink.api.common.functions.MapFunction;
import org.apache.flink.api.java.tuple.Tuple;
import org.apache.flink.api.java.tuple.Tuple2;
import org.apache.flink.streaming.api.datastream.DataStream;
import org.apache.flink.streaming.api.datastream.DataStreamSource;
import org.apache.flink.streaming.api.datastream.KeyedStream;
import org.apache.flink.streaming.api.datastream.SingleOutputStreamOperator;
import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
import org.apache.flink.util.Collector;

public class StreamWordCount {
    public static void main(String[] args) throws Exception {
        // 创建一个Stream计算执行环境
        StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
        // 调用Source创建DataStream
        DataStreamSource<String> lines = env.socketTextStream(args[0], Integer.parseInt(args[1]));
        int parallelism = lines.getParallelism();

//        DataStream<String> words = lines.flatMap(new FlatMapFunction<String, String>() {
//            @Override
//            public void flatMap(String line, Collector<String> out) throws Exception {
//                String[] words = line.split(" ");
//                for (String word : words) {
//                    out.collect(word);
//                }
//            }
//        });
//        SingleOutputStreamOperator<Tuple2<String, Integer>> wordAndOne = words.map(new MapFunction<String, Tuple2<String, Integer>>() {
//            @Override
//            public Tuple2<String, Integer> map(String word) throws Exception {
//                return Tuple2.of(word, 1);
//            }
//        });
        SingleOutputStreamOperator<Tuple2<String, Integer>> wordAndOne = lines.flatMap(new FlatMapFunction<String, Tuple2<String, Integer>>() {
            @Override
            public void flatMap(String line, Collector<Tuple2<String, Integer>> out) throws Exception {

                String[] words = line.split(" ");
                for (String word : words) {
                    out.collect(Tuple2.of(word, 1));
                }
            }
        });
        KeyedStream<Tuple2<String, Integer>, Tuple> keyed = wordAndOne.keyBy(0);
        SingleOutputStreamOperator<Tuple2<String, Integer>> summed = keyed.sum(1);
        //Transformation 结束
        //调用Sink
        summed.print();
        //执行程序
        env.execute("StreamWordCount");
    }
}

LambdaStreamWordCount（lambda的形式）

package cn._51doit.flink.day01;

import org.apache.flink.api.common.typeinfo.Types;
import org.apache.flink.api.java.tuple.Tuple2;
import org.apache.flink.streaming.api.datastream.DataStreamSource;
import org.apache.flink.streaming.api.datastream.SingleOutputStreamOperator;
import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
import org.apache.flink.util.Collector;

import java.util.Arrays;

public class LambdaStreamWordCount {
    public static void main(String[] args) throws Exception {
        // 创建一个stream计算的执行环境
        StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
        DataStreamSource<String> lines = env.socketTextStream("feng05", 8888);
//        SingleOutputStreamOperator<String> words = lines
//                .flatMap((String line, Collector<String> out) -> Arrays.asList(line.split(" ")).forEach(out::collect))
//                .returns(Types.STRING);
//
//        SingleOutputStreamOperator<Tuple2<String, Integer>> wordAndOne = words
//                .map(word -> Tuple2.of(word, 1))
//                .returns(Types.TUPLE(Types.STRING, Types.INT));
        SingleOutputStreamOperator<Tuple2<String, Integer>> wordAndOne = lines.flatMap((String line, Collector<Tuple2<String, Integer>> out)->{
            Arrays.stream(line.split(" ")).forEach(w -> out.collect(Tuple2.of(w, 1)));
        }).returns(Types.TUPLE(Types.STRING, Types.INT));
        SingleOutputStreamOperator<Tuple2<String, Integer>> result = wordAndOne.keyBy(0).sum(1);
        result.print();
        env.execute();
    }
}

自己运行时遇到的小问题

5.source

• 单并行source　

只有一个source来产生数据，如fromCollection、socketTextStream

• 双并行source

有多个source实例来产生数据

6 常用算子

6.1 keyBy

•  分组的对象是元组中的数据，可以直接指定角标，而且可以是多个

package cn._51doit.flink.day01;

import org.apache.flink.api.common.functions.MapFunction;
import org.apache.flink.api.java.tuple.Tuple;
import org.apache.flink.api.java.tuple.Tuple3;
import org.apache.flink.streaming.api.datastream.DataStreamSource;
import org.apache.flink.streaming.api.datastream.KeyedStream;
import org.apache.flink.streaming.api.datastream.SingleOutputStreamOperator;
import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;

public class KeyByDemo1 {

    public static void main(String[] args) throws Exception{

        StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();

        DataStreamSource<String> lines = env.socketTextStream("localhost", 8888);

        //辽宁省,沈阳市,1000
        SingleOutputStreamOperator<Tuple3<String, String, Double>> provinceCityAndMoney = lines.map(new MapFunction<String, Tuple3<String, String, Double>>() {

            @Override
            public Tuple3<String, String, Double> map(String line) throws Exception {
                String[] fields = line.split(",");
                String province = fields[0];
                String city = fields[1];
                double money = Double.parseDouble(fields[2]);
                return Tuple3.of(province, city, money);
            }
        });

        KeyedStream<Tuple3<String, String, Double>, Tuple> keyed = provinceCityAndMoney.keyBy(0, 1);

        SingleOutputStreamOperator<Tuple3<String, String, Double>> summed = keyed.sum(2);

        summed.print();

        env.execute();

    }
}

• 分组的对象不是元组中的元素，比如javabean中定义的字段，这个时候只能按照一个字段分组

package cn._51doit.flink.day01;

import org.apache.flink.api.common.functions.MapFunction;
import org.apache.flink.api.java.tuple.Tuple;
import org.apache.flink.api.java.tuple.Tuple3;
import org.apache.flink.streaming.api.datastream.DataStreamSource;
import org.apache.flink.streaming.api.datastream.KeyedStream;
import org.apache.flink.streaming.api.datastream.SingleOutputStreamOperator;
import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;

public class KeyByDemo2 {

    public static void main(String[] args) throws Exception{

        StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();

        DataStreamSource<String> lines = env.socketTextStream("localhost", 8888);

        //辽宁省,沈阳市,1000

        SingleOutputStreamOperator<OrderBean> provinceCityAndMoney = lines.map(new MapFunction<String, OrderBean>() {

            @Override
            public OrderBean map(String line) throws Exception {
                String[] fields = line.split(",");
                String province = fields[0];
                String city = fields[1];
                double money = Double.parseDouble(fields[2]);
                return new OrderBean(province, city, money);
            }
        });

        KeyedStream<OrderBean, Tuple> keyed = provinceCityAndMoney.keyBy("province", "city");

        SingleOutputStreamOperator<OrderBean> res = keyed.sum("money");

        //provinceCityAndMoney.keyBy(OrderBean::getProvince) 只能按照一个字段分组

        res.print();

        env.execute();

    }
}

6.2 max和min

　min、max返回分组的字段和参与比较的数据，如果有多个字段，其他字段的返回值是第一次出现的数据。

package cn._51doit.flink.day01;

import org.apache.flink.api.common.typeinfo.Types;
import org.apache.flink.api.java.tuple.Tuple;
import org.apache.flink.api.java.tuple.Tuple3;
import org.apache.flink.streaming.api.datastream.DataStreamSource;
import org.apache.flink.streaming.api.datastream.KeyedStream;
import org.apache.flink.streaming.api.datastream.SingleOutputStreamOperator;
import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;

public class MinMaxDemo {

    public static void main(String[] args) throws Exception {


        StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
        DataStreamSource<String> lines = env.socketTextStream("localhost", 8888);

        //省份,城市,人数
        SingleOutputStreamOperator<Tuple3<String, String, Integer>> provinceCityAmount = lines.map(line -> {
            String[] fields = line.split(",");
            String province = fields[0];
            String city = fields[1];
            Integer amount = Integer.parseInt(fields[2]);
            return Tuple3.of(province, city, amount);
        }).returns(Types.TUPLE(Types.STRING, Types.STRING, Types.INT));

        KeyedStream<Tuple3<String, String, Integer>, Tuple> keyed = provinceCityAmount.keyBy(0);

        //min、max返回分组的字段和参与比较的数据，如果有多个字段，其他字段的返回值是第一次出现的数据。
        SingleOutputStreamOperator<Tuple3<String, String, Integer>> max = keyed.max(2);

        max.print();

        env.execute();
    }

}

比如

江西,鹰潭,1000   //先输入此数据，max后得到本身
江西,南昌,2000　 //输入该数据，max后得到的是江西,鹰潭,2000  并不能得到南昌字段

解决办法=====>使用maxBy和minBy

6.2 maxBy和minBy

package cn._51doit.flink.day01;

import org.apache.flink.api.common.typeinfo.Types;
import org.apache.flink.api.java.tuple.Tuple;
import org.apache.flink.api.java.tuple.Tuple3;
import org.apache.flink.streaming.api.datastream.DataStreamSource;
import org.apache.flink.streaming.api.datastream.KeyedStream;
import org.apache.flink.streaming.api.datastream.SingleOutputStreamOperator;
import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;

public class MinByMaxByDemo {

    public static void main(String[] args) throws Exception {


        StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
        DataStreamSource<String> lines = env.socketTextStream("localhost", 8888);

        //省份,城市,人数
        江西,鹰潭,1000
        江西,南昌,2000
        SingleOutputStreamOperator<Tuple3<String, String, Integer>> provinceCityAmount = lines.map(line -> {
            String[] fields = line.split(",");
            String province = fields[0];
            String city = fields[1];
            Integer amount = Integer.parseInt(fields[2]);
            return Tuple3.of(province, city, amount);
        }).returns(Types.TUPLE(Types.STRING, Types.STRING, Types.INT));

        KeyedStream<Tuple3<String, String, Integer>, Tuple> keyed = provinceCityAmount.keyBy(0);

        //minBy、maxBy返回最大值或最小值数据本身（全部字段都返回）。
        SingleOutputStreamOperator<Tuple3<String, String, Integer>> max = keyed.maxBy(2);

        max.print();

        env.execute();
    }

}

这种形式又会出现另外一个难点，就是当按照key进行分组后，比较大小的值一样时，其它字段返回的值又是第一次出现的数据，解决办法===>加一个参数（可以从源码中得出此结论），如下：加上false

SingleOutputStreamOperator<Tuple3<String, String, Integer>> max = keyed.maxBy(2, false);

此时其它字段返回的值就是最后依次出现的字段了。