广播流是什么?
将一条数据广播到所有的节点。使用 dataStream.broadCast()
广播流使用场景?
一般用于动态加载配置项。比如lol,每天不断有人再投诉举报,客服根本忙不过来,腾讯内部做了一个判断,只有vip3以上的客户的投诉才会有人工一对一回复,过了一段时间大家都发现vip3才有人工,都开始充钱到vip3,此时人还是很多,于是只有vip4上的客户才能人工回复
vip3->vip4 这种判断标准在不断的变化。此时就需要广播流。因为此时数据只有1条,需要多个节点都收到这个变化的数据。
广播流怎么用?
一般通过connect合流去操作 a connect b.broadcast 。a是主流也就是数据流,b是配置变化流
不多说直接上demo,开箱即用
package com.chenchi.broadcast;
import org.apache.flink.api.common.state.BroadcastState;
import org.apache.flink.api.common.state.MapStateDescriptor;
import org.apache.flink.api.common.state.ValueState;
import org.apache.flink.api.common.state.ValueStateDescriptor;
import org.apache.flink.api.common.typeinfo.Types;
import org.apache.flink.configuration.Configuration;
import org.apache.flink.streaming.api.datastream.BroadcastStream;
import org.apache.flink.streaming.api.datastream.DataStream;
import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
import org.apache.flink.streaming.api.functions.co.CoProcessFunction;
import org.apache.flink.streaming.api.functions.co.KeyedBroadcastProcessFunction;
import org.apache.flink.streaming.api.functions.source.SourceFunction;
import org.apache.flink.util.Collector;
import java.util.HashMap;
import java.util.Random;
public class BroadCastStreamDemo {
public static void main(String[] args) throws Exception {
StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
DataStream patternDataStream = env.addSource(new ChangeSource());
DataStream userDataStream = env.addSource(new CustomerSource());
userDataStream.print("user");
patternDataStream.print("pattern");
//test1 直接合流 不广播。只会在一个节点更新。 用于特殊需求?
// userDataStream
// .keyBy(user -> user.userId)
// .connect(patternDataStream)
// .process(new CustomerSimpleProcess())
// .print();
//test2
// 定义广播状态的描述器,创建广播流 如何保存需要的广播数据呢 这个案例是通过map保留变化数据
// 服务器托管网 userDataStream
// .keyBy(user -> user.userId)
// .connect(patternDataStream.broadcast())
// .process(new CustomerSimpleProcess())
// .print();
//test3
MapStateDescriptor bcStateDescriptor = new MapStateDescriptor(
"patterns", Types.VOID, Types.POJO(Pattern.class));
//通过描述器 更新
BroadcastStream broadcast = patternDataStream.broadcast(bcStateDescriptor);
userDataStream
.keyBy(user -> user.userId)
.connect(broadcast)
.process(new CustomerBroadCastProcess())
.print();
env.execute();
}
private static服务器托管网 class CustomerBroadCastProcess extends KeyedBroadcastProcessFunction {
@Override
public void processElement(User user, KeyedBroadcastProcessFunction.ReadOnlyContext readOnlyContext, Collector collector) throws Exception {
Integer userVip = user.getVip();
//获取广播流的数据 不是通过map保存
Pattern pattern = readOnlyContext.getBroadcastState(new MapStateDescriptor("patterns", Types.VOID, Types.POJO(Pattern.class))).get(null);
if (pattern!=null){
Integer patternVip = pattern.vip;
String result = "当前系统需要的vip等级=" + patternVip + ",用户id=" + user.userId + ",vip=" + userVip;
if (userVip>= patternVip){
result=result+"符合要求";
}else {
result=result+"不符合要求";
}
collector.collect(result);
}else {
System.out.println("pattern is null ");
}
}
@Override
public void processBroadcastElement(Pattern pattern, KeyedBroadcastProcessFunction.Context context, Collector collector) throws Exception {
BroadcastState bcState = context.getBroadcastState(
new MapStateDescriptor("patterns", Types.VOID, Types.POJO(Pattern.class)));
// 将广播状态更新为当前的pattern
bcState.put(null, pattern);
}
}
public static class CustomerSimpleProcess extends CoProcessFunction {
ValueState vip; //这个是保留主流的state的。 不是保留广播流的state
HashMap vipMap;
@Override
public void open(Configuration parameters) throws Exception {
vip = getRuntimeContext().getState(new ValueStateDescriptor("vip", Integer.class));
vipMap=new HashMap();
super.open(parameters);
}
@Override
public void processElement1(User user, CoProcessFunction.Context context, Collector collector) throws Exception {
Integer userVip = user.getVip();
Integer patternVip = vipMap.getOrDefault("vip", 0);
String result = "当前系统需要的vip等级=" + patternVip + ",用户id=" + user.userId + ",vip=" + userVip;
if (userVip>=patternVip){
result=result+"符合要求";
}else {
result=result+"不符合要求";
}
collector.collect(result);
}
@Override
public void processElement2(Pattern pattern, CoProcessFunction.Context context, Collector collector) throws Exception {
vipMap.put("vip",pattern.vip);
}
}
public static class User {
public Integer userId;
public Integer vip;
public User() {
}
public User(Integer userId, Integer vip) {
this.userId = userId;
this.vip = vip;
}
public Integer getUserId() {
return userId;
}
public void setUserId(Integer userId) {
this.userId = userId;
}
public Integer getVip() {
return vip;
}
public void setVip(Integer vip) {
this.vip = vip;
}
@Override
public String toString() {
return "Action{" +
"userId=" + userId +
", vip='" + vip + ''' +
'}';
}
}
// 定义行为模式POJO类,包含先后发生的两个行为
public static class Pattern {
public Integer vip;
public Pattern() {
}
public Pattern(Integer vip) {
this.vip = vip;
}
@Override
public String toString() {
return "Pattern{" +
"vip='" + vip + ''' +
'}';
}
}
private static class CustomerSource implements SourceFunction {
boolean run = true;
@Override
public void run(SourceContext sourceContext) throws Exception {
while (true) {
Integer userId = new Random().nextInt(1000);
Integer vip = new Random().nextInt(10);
sourceContext.collect(new User(userId, vip));
Thread.sleep(1000);
}
}
@Override
public void cancel() {
run = false;
}
}
private static class ChangeSource implements SourceFunction {
boolean run = true;
@Override
public void run(SourceContext sourceContext) throws Exception {
int i = 1;
while (true) {
sourceContext.collect(new Pattern(i++));
Thread.sleep(5000);
}
}
@Override
public void cancel() {
run = false;
}
}
}
demo思想:以上述vip做例子,获取用户不断投诉的id和vip等级, 数据库保存可以享受人工服务的vip等级,该等级可以自行调整(我是随着时间变化主键增大)。
test1 不广播
注意看pattern:4 print vip=2的消息但是不代表是task4收到的消息,我们看到>1输出了vip=2
但是task10 task9都还是vip=0 ,说明流没有广播,除非此处并行度设置为1
test2 map保存变化数据
test3通过描述器获取数据
和test2 一样,不过要注意因为两个流的数据有先后,可能还没有pattern就来了user信息,所以建议先初始化,或者先添加pattern流。
服务器托管,北京服务器托管,服务器租用 http://www.fwqtg.net
前言 容器是一种虚拟化技术,用于封装和运行应用程序及其依赖项,以便在不同的计算环境中保持一致性和可移植性。自2013年容器诞生至今,容器Docker镜像的下载量超20亿,虽然容器行业发展如火如荼,但是其安全风险却不容乐观, 据《Sysdig 2022 云原生安…