状态分类:算子状态(Operatior State);键控状态(Keyed State);状态后端(StateBackends)
什么是状态
1:由一个任务维护,并且用来计算某个结果的所有数据,都属于这个任务的状态
2:可以认为状态就是一个本地变量,可以被任务的业务逻辑访问
3:Flink 会进行状态管理,包括状态一致性、故障处理以及高效存储和访问,以便开发人员可以专注于应用程序的逻辑
算子状态(Operator State)
1:算子状态的作用范围限定为算子任务,由同一并行任务所处理的所有数据都可以访问到相同的状态
2:状态对于同一子任务而言是共享的
3: 算子状态不能由相同或不同算子的另一个子任务访问
列表状态(List state):将状态表示为一组数据的列表
联合列表状态(Union list state) :也将状态表示为数据的列表。它与常规列表状态的区别在于,在发生故障时,或者从保存点(savepoint)启动应用程序时如何恢复
广播状态(Broadcast state) :如果一个算子有多项任务,而它的每项任务状态又都相同,那么这种特殊情况最适合应用广播状态。
class MyCountMap implements MapFunction<sensorreading, tuple2<string,integer>>, ListCheckpointed<integer> {
//定义算子状态
private Integer count = 0;
@Override
public Tuple2<string, integer> map(SensorReading value) throws Exception {
count++;
return new Tuple2<>(value.getId(),count);
}
@Override
public List<integer> snapshotState(long checkpointId, long timestamp) throws Exception {
return Collections.singletonList(count);
}
@Override
public void restoreState(List<integer> state) throws Exception {
for (Integer integer : state) {
count+=integer;
}
}
}
</integer></integer></string,></integer></sensorreading,>
键控状态(Keyed State)
1:键控状态是根据输入数据流中定义的键(key)来维护和访问的
2:Flink 为每个 key 维护一个状态实例,并将具有相同键的所有数据,都分区到
3:同一个算子任务中,这个任务会维护和处理这个 key 对应的状态
4:当任务处理一条数据时,它会自动将状态的访问范围限定为当前数据的 key
1:值状态(Value state) :将状态表示为单个的值
2:列表状态(List state) :将状态表示为一组数据的列表 3:映射状态(Map state)将状态表示为一组 Key-Value 对
4:聚合状态(Reducing state & Aggregating State) :将状态表示为一个用于聚合操作的列表
package com.atguigu.state;
import com.atguigu.bean.SensorReading;
import org.apache.flink.api.common.functions.MapFunction;
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.configuration.Configuration;
import org.apache.flink.streaming.api.checkpoint.ListCheckpointed;
import org.apache.flink.streaming.api.datastream.DataStream;
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 scala.Tuple2;
import java.util.Collections;
import java.util.List;
public class KeyedState {
public static void main(String[] args) throws Exception {
StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
env.setParallelism(1);
DataStreamSource<string> socketTextStream = env.socketTextStream("hadoop112", 7777);
DataStream<sensorreading> inputStream = socketTextStream.map(value -> {
String[] split = value.split(",");
return new SensorReading(split[0], new Long(split[1]), new Double(split[2]));
});
/**
* 统计每个编号的温度个数
*/
SingleOutputStreamOperator<tuple2<string,integer>> map = inputStream.keyBy("id").map(new MyCountMap2());
map.print();
env.execute();
}
}
class MyCountMap2 extends RichMapFunction<sensorreading, tuple2<string,integer>> {
//定义算子状态
private ValueState<integer> myKed ;
@Override
public void open(Configuration parameters) throws Exception {
myKed = getRuntimeContext().getState(new ValueStateDescriptor<integer>("key_count",Integer.class,0));
}
@Override
public Tuple2<string,integer> map(SensorReading value) throws Exception {
Integer count = myKed.value();
count++;
myKed.update(count);
return new Tuple2<>(value.getId(),count);
}
}
</string,integer></integer></integer></sensorreading,></tuple2<string,integer></sensorreading></string>
状态后端(State Backends)
1:每传入一条数据,有状态的算子任务都会读取和更新状态
2:由于有效的状态访问对于处理数据的低延迟至关重要,因此每个并行任务都会在本地维护其状态,以确保快速的状态访问 状态的存储、访问以及维护,由一个可插入的组件决定,这个组件就
叫做 状态后端 (state backend)
3:状态后端主要负责两件事:本地的状态管理,以及将检查点(checkpoint)状态写入远程存储
选择一个状态后端
➢ MemoryStateBackend
• 内存级的状态后端,会将键控状态作为内存中的对象进行管理,将它们存储在TaskManager 的 JVM 堆上,而将 checkpoint 存储在 JobManager 的内存中
• 特点:快速、低延迟,但不稳定
➢ FsStateBackend
• 将 checkpoint 存到远程的持久化文件系统(FileSystem)上,而对于本地状态,跟 MemoryStateBackend 一样,也会存在 TaskManager 的 JVM 堆上同时拥有内存级的本地访问速度,和更好的容错保证
➢ RocksDBStateBackend
• 将所有状态序列化后,存入本地的 RocksDB 中存储。
状态编程
同一个传感器前后温度相差10就报警输出:(sensor_1,35.8,50.0)
package com.atguigu.state;
import com.atguigu.bean.SensorReading;
import org.apache.flink.api.common.functions.RichFlatMapFunction;
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.Tuple3;
import org.apache.flink.configuration.Configuration;
import org.apache.flink.streaming.api.datastream.DataStream;
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 scala.Tuple2;
public class KeyedStateTest {
public static void main(String[] args) throws Exception {
StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
env.setParallelism(1);
DataStreamSource<string> socketTextStream = env.socketTextStream("hadoop112", 7777);
DataStream<sensorreading> inputStream = socketTextStream.map(value -> {
String[] split = value.split(",");
return new SensorReading(split[0], new Long(split[1]), new Double(split[2]));
});
/**
* 统计每个编号的温度个数
*/
SingleOutputStreamOperator<tuple3<string,double,double>> map = inputStream.keyBy("id").flatMap(new MyFlatMap(10.0));
map.print();
env.execute();
}
}
class MyFlatMap extends RichFlatMapFunction<sensorreading,tuple3<string,double,double>> {
//当前温度跳变差值
private double temp;
public MyFlatMap(double d){
this.temp = d;
}
private ValueState<double> lastTempState = null;
@Override
public void open(Configuration parameters) throws Exception {
lastTempState = getRuntimeContext().getState(new ValueStateDescriptor<double>("last_State",Double.class));
}
@Override
public void flatMap(SensorReading value, Collector<tuple3<string, double, double>> out) throws Exception {
//获取上次的温度值
Double lastTemp = lastTempState.value();
if(lastTemp!=null){
Double diff = Math.abs(value.getTemperature()-lastTemp);
if(diff>=temp){
out.collect(new Tuple3<>(value.getId(),lastTemp,value.getTemperature()));
}
}
//更新状态
lastTempState.update(value.getTemperature());
}
@Override
public void close() throws Exception {
lastTempState.clear();
}
}
</tuple3<string,></double></double></sensorreading,tuple3<string,double,double></tuple3<string,double,double></sensorreading></string>
Original: https://blog.csdn.net/m0_57126456/article/details/121502871
Author: 天才少年137
Title: Flink——状态管理