15-State-理解原理即可

State-理解原理即可

Flink中状态的自动管理

之前写的Flink代码中其实已经做好了状态自动管理,如

发送hello ,得出(hello,1)

再发送hello ,得出(hello,2)

说明Flink已经自动的将当前数据和历史状态/历史结果进行了聚合,做到了状态的自动管理

在实际开发中绝大多数情况下,我们直接使用自动管理即可

一些特殊情况才会使用手动的状态管理!—后面项目中会使用!

所以这里得先学习state状态如何手动管理!

package cn.itcast.source;

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

/**
 * Author itcast
 * Desc 演示DataStream-Source-基于Socket
 */
public class SourceDemo03_Socket {
    public static void main(String[] args) throws Exception {
        //TODO 0.env
        StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
        env.setRuntimeMode(RuntimeExecutionMode.AUTOMATIC);

        //TODO 1.source
        DataStream<String> lines = env.socketTextStream("node1", 9999);


        //TODO 2.transformation
        /*SingleOutputStreamOperator<String> words = lines.flatMap(new FlatMapFunction<String, String>() {
            @Override
            public void flatMap(String value, Collector<String> out) throws Exception {
                String[] arr = value.split(" ");
                for (String word : arr) {
                    out.collect(word);
                }
            }
        });

        words.map(new MapFunction<String, Tuple2<String,Integer>>() {
            @Override
            public Tuple2<String, Integer> map(String value) throws Exception {
                return Tuple2.of(value,1);
            }
        });*/

        //注意:下面的操作将上面的2步合成了1步,直接切割单词并记为1返回
        SingleOutputStreamOperator<Tuple2<String, Integer>> wordAndOne = lines.flatMap(new FlatMapFunction<String, Tuple2<String, Integer>>() {
            @Override
            public void flatMap(String value, Collector<Tuple2<String, Integer>> out) throws Exception {
                String[] arr = value.split(" ");
                for (String word : arr) {
                    out.collect(Tuple2.of(word, 1));
                }
            }
        });

        SingleOutputStreamOperator<Tuple2<String, Integer>> result = wordAndOne.keyBy(t -> t.f0).sum(1);

        //TODO 3.sink
        result.print();

        //TODO 4.execute
        env.execute();
    }
}

无状态计算和有状态计算

• 无状态计算,不需要考虑历史值, 如map

hello --> (hello,1)

hello --> (hello,1)

[外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传(img-56Sbptxv-1624947045215)(Flink-day03.assets/1610871016671.png)]

• 有状态计算,需要考虑历史值,如:sum

hello , (hello,1)

hello , (hello,2)

状态分类

• State
  • ManagerState–开发中推荐使用 : Fink自动管理/优化,支持多种数据结构
    • KeyState–只能在keyedStream上使用,支持多种数据结构
      • OperatorState–一般用在Source上,支持ListState
  • RawState–完全由用户自己管理,只支持byte[],只能在自定义Operator上使用
    • OperatorState

分类详细图解:

代码演示-ManagerState-keyState

https://ci.apache.org/projects/flink/flink-docs-release-1.12/dev/stream/state/state.html

package cn.itcast.state;

import org.apache.flink.api.common.RuntimeExecutionMode;
import org.apache.flink.api.common.functions.RichMapFunction;
import org.apache.flink.api.common.state.ValueState;
import org.apache.flink.api.common.state.ValueStateDescriptor;
import org.apache.flink.api.java.tuple.Tuple2;
import org.apache.flink.api.java.tuple.Tuple3;
import org.apache.flink.configuration.Configuration;
import org.apache.flink.streaming.api.datastream.DataStream;
import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;

/**
 * Author itcast
 * Desc 使用KeyState中的ValueState获取流数据中的最大值/实际中可以使用maxBy即可
 */
public class StateDemo01_KeyState {
    public static void main(String[] args) throws Exception {
        //TODO 0.env
        StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
        env.setRuntimeMode(RuntimeExecutionMode.AUTOMATIC);

        //TODO 1.source
        DataStream<Tuple2<String, Long>> tupleDS = env.fromElements(
                Tuple2.of("北京", 1L),
                Tuple2.of("上海", 2L),
                Tuple2.of("北京", 6L),
                Tuple2.of("上海", 8L),
                Tuple2.of("北京", 3L),
                Tuple2.of("上海", 4L)
        );

        //TODO 2.transformation
        //需求:求各个城市的value最大值
        //实际中使用maxBy即可
        DataStream<Tuple2<String, Long>> result1 = tupleDS.keyBy(t -> t.f0).maxBy(1);

        //学习时可以使用KeyState中的ValueState来实现maxBy的底层
        DataStream<Tuple3<String, Long, Long>> result2 = tupleDS.keyBy(t -> t.f0).map(new RichMapFunction<Tuple2<String, Long>, Tuple3<String, Long, Long>>() {
            //-1.定义一个状态用来存放最大值
            private ValueState<Long> maxValueState;

            //-2.状态初始化
            @Override
            public void open(Configuration parameters) throws Exception {
                //创建状态描述器
                ValueStateDescriptor stateDescriptor = new ValueStateDescriptor("maxValueState", Long.class);
                //根据状态描述器获取/初始化状态
                maxValueState = getRuntimeContext().getState(stateDescriptor);
            }

            //-3.使用状态
            @Override
            public Tuple3<String, Long, Long> map(Tuple2<String, Long> value) throws Exception {
                Long currentValue = value.f1;
                //获取状态
                Long historyValue = maxValueState.value();
                //判断状态
                if (historyValue == null || currentValue > historyValue) {
                    historyValue = currentValue;
                    //更新状态
                    maxValueState.update(historyValue);
                    return Tuple3.of(value.f0, currentValue, historyValue);
                } else {
                    return Tuple3.of(value.f0, currentValue, historyValue);
                }
            }
        });


        //TODO 3.sink
        //result1.print();
        //4> (北京,6)
        //1> (上海,8)
        result2.print();
        //1> (上海,xxx,8)
        //4> (北京,xxx,6)

        //TODO 4.execute
        env.execute();
    }
}

代码演示-ManagerState-OperatorState

https://ci.apache.org/projects/flink/flink-docs-release-1.12/dev/stream/state/state.html

package cn.itcast.state;

import org.apache.flink.api.common.RuntimeExecutionMode;
import org.apache.flink.api.common.restartstrategy.RestartStrategies;
import org.apache.flink.api.common.state.ListState;
import org.apache.flink.api.common.state.ListStateDescriptor;
import org.apache.flink.runtime.state.FunctionInitializationContext;
import org.apache.flink.runtime.state.FunctionSnapshotContext;
import org.apache.flink.runtime.state.filesystem.FsStateBackend;
import org.apache.flink.streaming.api.CheckpointingMode;
import org.apache.flink.streaming.api.checkpoint.CheckpointedFunction;
import org.apache.flink.streaming.api.datastream.DataStreamSource;
import org.apache.flink.streaming.api.environment.CheckpointConfig;
import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
import org.apache.flink.streaming.api.functions.source.RichParallelSourceFunction;

import java.util.Iterator;

/**
 * Author itcast
 * Desc 使用OperatorState中的ListState模拟KafkaSource进行offset维护
 */
public class StateDemo02_OperatorState {
    public static void main(String[] args) throws Exception {
        //TODO 0.env
        StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
        env.setRuntimeMode(RuntimeExecutionMode.AUTOMATIC);
        env.setParallelism(1);//并行度设置为1方便观察
        //下面的Checkpoint和重启策略配置先直接使用,下次课学
        env.enableCheckpointing(1000);//每隔1s执行一次Checkpoint
        env.setStateBackend(new FsStateBackend("file:///D:/ckp"));
        env.getCheckpointConfig().enableExternalizedCheckpoints(CheckpointConfig.ExternalizedCheckpointCleanup.RETAIN_ON_CANCELLATION);
        env.getCheckpointConfig().setCheckpointingMode(CheckpointingMode.EXACTLY_ONCE);
        //固定延迟重启策略: 程序出现异常的时候，重启2次，每次延迟3秒钟重启，超过2次，程序退出
        env.setRestartStrategy(RestartStrategies.fixedDelayRestart(2, 3000));

        //TODO 1.source
        DataStreamSource<String> ds = env.addSource(new MyKafkaSource()).setParallelism(1);

        //TODO 2.transformation

        //TODO 3.sink
        ds.print();

        //TODO 4.execute
        env.execute();
    }
    //使用OperatorState中的ListState模拟KafkaSource进行offset维护
    public static class MyKafkaSource extends RichParallelSourceFunction<String> implements CheckpointedFunction {
        private boolean flag = true;
        //-1.声明ListState
        private ListState<Long> offsetState = null; //用来存放offset
        private Long offset = 0L;//用来存放offset的值

        //-2.初始化/创建ListState
        @Override
        public void initializeState(FunctionInitializationContext context) throws Exception {
            ListStateDescriptor<Long> stateDescriptor = new ListStateDescriptor<>("offsetState", Long.class);
            offsetState = context.getOperatorStateStore().getListState(stateDescriptor);
        }
        //-3.使用state
        @Override
        public void run(SourceContext<String> ctx) throws Exception {
            while (flag){
                Iterator<Long> iterator = offsetState.get().iterator();
                if(iterator.hasNext()){
                    offset = iterator.next();
                }
                offset += 1;
                int subTaskId = getRuntimeContext().getIndexOfThisSubtask();
                ctx.collect("subTaskId:"+ subTaskId + ",当前的offset值为:"+offset);
                Thread.sleep(1000);

                //模拟异常
                if(offset % 5 == 0){
                    throw new Exception("bug出现了.....");
                }
            }
        }
        //-4.state持久化
        //该方法会定时执行将state状态从内存存入Checkpoint磁盘目录中
        @Override
        public void snapshotState(FunctionSnapshotContext context) throws Exception {
            offsetState.clear();//清理内容数据并存入Checkpoint磁盘目录中
            offsetState.add(offset);
        }

        @Override
        public void cancel() {
            flag = false;
        }
    }
}