最近看了大佬的博客，突然想起Async I/O方式是Blink 推给社区的一大重要功能，可以使用异步的方式获取外部数据，想着自己实现以下，项目上用的时候，可以不用现去找了。

最开始想用scala 实现一个读取 hbase数据的demo，参照官网demo：

/*** An implementation of the 'AsyncFunction' that sends requests and sets the callback.*/

class AsyncDatabaseRequest extendsAsyncFunction[String, (String, String)] {/**The database specific client that can issue concurrent requests with callbacks*/lazy val client: DatabaseClient= newDatabaseClient(host, post, credentials)/**The context used for the future callbacks*/implicit lazy val executor: ExecutionContext=ExecutionContext.fromExecutor(Executors.directExecutor())

override def asyncInvoke(str: String, resultFuture: ResultFuture[(String, String)]): Unit={//issue the asynchronous request, receive a future for the result

val resultFutureRequested: Future[String] = client.query(str)//set the callback to be executed once the request by the client is complete//the callback simply forwards the result to the result future

resultFutureRequested.onSuccess {case result: String =>resultFuture.complete(Iterable((str, result)))

}

}

}//create the original stream

val stream: DataStream[String] =...//apply the async I/O transformation

val resultStream: DataStream[(String, String)] =AsyncDataStream.unorderedWait(stream, new AsyncDatabaseRequest(), 1000, TimeUnit.MILLISECONDS, 100)

失败了，上图标红的部分实现不了

1、Future 找不到可以用的实现类

2、unorderedWait 一直报错

源码example 里面也有Scala 的案例

def main(args: Array[String]) {

val timeout= 10000Lval env=StreamExecutionEnvironment.getExecutionEnvironment

val input= env.addSource(newSimpleSource())

val asyncMapped= AsyncDataStream.orderedWait(input, timeout, TimeUnit.MILLISECONDS, 10) {

(input, collector: ResultFuture[Int]) =>Future {

collector.complete(Seq(input))

} (ExecutionContext.global)

}

asyncMapped.print()

env.execute("Async I/O job")

}

主要部分是这样的，菜鸡表示无力，想继承RichAsyncFunction，可以使用open 方法初始化链接。

网上博客翻了不少，大部分是翻译官网的原理，案例也没有可以执行的，苦恼。

失败了。

转为java版本的，昨天在群里问，有个大佬给我个Java版本的： https://github.com/perkinls/flink-local-train/blob/c8b4efe33620352aea0100adef4fae2a068a3b65/src/main/scala/com/lp/test/asyncio/AsyncIoSideTableJoinMysqlJava.java 还没看过，因为Java版的官网的案例能看懂。

下面开始上mysql 版本 的 源码(hbase 的还没测试过，本机的hbase 挂了)：

业务如下：

接收kafka数据，转为user对象，调用async，使用user.id 查询对应的phone，放回user对象，输出

主类：

importcom.alibaba.fastjson.JSON;importcom.venn.common.Common;importorg.apache.flink.formats.json.JsonNodeDeserializationSchema;importorg.apache.flink.shaded.jackson2.com.fasterxml.jackson.databind.node.ObjectNode;importorg.apache.flink.streaming.api.datastream.AsyncDataStream;importorg.apache.flink.streaming.api.datastream.DataStream;importorg.apache.flink.streaming.api.environment.StreamExecutionEnvironment;importorg.apache.flink.streaming.connectors.kafka.FlinkKafkaConsumer;importjava.util.concurrent.TimeUnit;public classAsyncMysqlRequest {public static void main(String[] args) throwsException {final StreamExecutionEnvironment env =StreamExecutionEnvironment.getExecutionEnvironment();

FlinkKafkaConsumer source = new FlinkKafkaConsumer<>("async", newJsonNodeDeserializationSchema(), Common.getProp());//接收kafka数据，转为User 对象

DataStream input = env.addSource(source).map(value ->{

String id= value.get("id").asText();

String username= value.get("username").asText();

String password= value.get("password").asText();return newUser(id, username, password);

});//异步IO 获取mysql数据, timeout 时间 1s，容量 10(超过10个请求，会反压上游节点)

DataStream async = AsyncDataStream.unorderedWait(input, new AsyncFunctionForMysqlJava(), 1000, TimeUnit.MICROSECONDS, 10);

async.map(user->{returnJSON.toJSON(user).toString();

})

.print();

env.execute("asyncForMysql");

}

}

函数类：

importorg.apache.flink.configuration.Configuration;importorg.apache.flink.streaming.api.functions.async.ResultFuture;importorg.apache.flink.streaming.api.functions.async.RichAsyncFunction;importorg.slf4j.Logger;importorg.slf4j.LoggerFactory;importjava.util.ArrayList;importjava.util.Collections;importjava.util.List;import java.util.concurrent.*;public class AsyncFunctionForMysqlJava extends RichAsyncFunction{

Logger logger= LoggerFactory.getLogger(AsyncFunctionForMysqlJava.class);private transientMysqlClient client;private transientExecutorService executorService;/*** open 方法中初始化链接

*

*@paramparameters

*@throwsException*/@Overridepublic void open(Configuration parameters) throwsException {

logger.info("async function for mysql java open ...");super.open(parameters);

client= newMysqlClient();

executorService= Executors.newFixedThreadPool(30);

}/*** use asyncUser.getId async get asyncUser phone

*

*@paramasyncUser

*@paramresultFuture

*@throwsException*/@Overridepublic void asyncInvoke(AsyncUser asyncUser, ResultFuture resultFuture) throwsException {

executorService.submit(()->{//submit query

System.out.println("submit query : " + asyncUser.getId() + "-1-" +System.currentTimeMillis());

AsyncUser tmp=client.query1(asyncUser);//一定要记得放回 resultFuture，不然数据全部是timeout 的

resultFuture.complete(Collections.singletonList(tmp));

});

}

@Overridepublic void timeout(AsyncUser input, ResultFuture resultFuture) throwsException {

logger.warn("Async function for hbase timeout");

List list = newArrayList();

input.setPhone("timeout");

list.add(input);

resultFuture.complete(list);

}/*** close function

*

*@throwsException*/@Overridepublic void close() throwsException {

logger.info("async function for mysql java close ...");super.close();

}

}

MysqlClient：

importcom.venn.flink.util.MathUtil;importorg.apache.flink.shaded.netty4.io.netty.channel.DefaultEventLoop;importorg.apache.flink.shaded.netty4.io.netty.util.concurrent.Future;importorg.apache.flink.shaded.netty4.io.netty.util.concurrent.SucceededFuture;importjava.sql.DriverManager;importjava.sql.PreparedStatement;importjava.sql.ResultSet;importjava.sql.SQLException;public classMysqlClient {private static String jdbcUrl = "jdbc:mysql://192.168.229.128:3306?useSSL=false&allowPublicKeyRetrieval=true";private static String username = "root";private static String password = "123456";private static String driverName = "com.mysql.jdbc.Driver";private staticjava.sql.Connection conn;private staticPreparedStatement ps;static{try{

Class.forName(driverName);

conn=DriverManager.getConnection(jdbcUrl, username, password);

ps= conn.prepareStatement("select phone from async.async_test where id = ?");

}catch (ClassNotFoundException |SQLException e) {

e.printStackTrace();

}

}/*** execute query

*

*@paramuser

*@return

*/

publicAsyncUser query1(AsyncUser user) {try{

Thread.sleep(10);

}catch(InterruptedException e) {

e.printStackTrace();

}

String phone= "0000";try{

ps.setString(1, user.getId());

ResultSet rs=ps.executeQuery();if (!rs.isClosed() &&rs.next()) {

phone= rs.getString(1);

}

System.out.println("execute query : " + user.getId() + "-2-" + "phone : " + phone + "-" +System.currentTimeMillis());

}catch(SQLException e) {

e.printStackTrace();

}

user.setPhone(phone);returnuser;

}

// 测试代码public static voidmain(String[] args) {

MysqlClient mysqlClient= newMysqlClient();

AsyncUser asyncUser= newAsyncUser();

asyncUser.setId("526");long start =System.currentTimeMillis();

asyncUser=mysqlClient.query1(asyncUser);

System.out.println("end : " + (System.currentTimeMillis() -start));

System.out.println(asyncUser.toString());

}

}

函数类(错误示范：asyncInvoke 方法中阻塞查询数据库，是同步的)：

importorg.apache.flink.configuration.Configuration;importorg.apache.flink.streaming.api.functions.async.ResultFuture;importorg.apache.flink.streaming.api.functions.async.RichAsyncFunction;importorg.slf4j.Logger;importorg.slf4j.LoggerFactory;importjava.sql.DriverManager;importjava.sql.PreparedStatement;importjava.sql.ResultSet;importjava.util.ArrayList;importjava.util.List;public class AsyncFunctionForMysqlJava extends RichAsyncFunction{//链接

private static String jdbcUrl = "jdbc:mysql://192.168.229.128:3306?useSSL=false";private static String username = "root";private static String password = "123456";private static String driverName = "com.mysql.jdbc.Driver";

java.sql.Connection conn;

PreparedStatement ps;

Logger logger= LoggerFactory.getLogger(AsyncFunctionForMysqlJava.class);/*** open 方法中初始化链接

*@paramparameters

*@throwsException*/@Overridepublic void open(Configuration parameters) throwsException {

logger.info("async function for hbase java open ...");super.open(parameters);

Class.forName(driverName);

conn=DriverManager.getConnection(jdbcUrl, username, password);

ps= conn.prepareStatement("select phone from async.async_test where id = ?");

}/*** use user.getId async get user phone

*

*@paramuser

*@paramresultFuture

*@throwsException*/@Overridepublic void asyncInvoke(User user, ResultFuture resultFuture) throwsException {//使用 user id 查询

ps.setString(1, user.getId());

ResultSet rs= ps.executeQuery();

String phone = null;

if (rs.next()) {

phone = rs.getString(1);

}

user.setPhone(phone);

List list = newArrayList();

list.add(user);//放回 result 队列

resultFuture.complete(list);

}

@Overridepublic void timeout(User input, ResultFuture resultFuture) throwsException {

logger.info("Async function for hbase timeout");

List list = newArrayList();

list.add(input);

resultFuture.complete(list);

}/*** close function

*

*@throwsException*/@Overridepublic void close() throwsException {

logger.info("async function for hbase java close ...");super.close();

conn.close();

}

}

测试数据如下：

{"id" : 1, "username" : "venn", "password" : 1561709530935}

{"id" : 2, "username" : "venn", "password" : 1561709536029}

{"id" : 3, "username" : "venn", "password" : 1561709541033}

{"id" : 4, "username" : "venn", "password" : 1561709546037}

{"id" : 5, "username" : "venn", "password" : 1561709551040}

{"id" : 6, "username" : "venn", "password" : 1561709556044}

{"id" : 7, "username" : "venn", "password" : 1561709561048}

执行结果如下：

submit query : 1-1-1562763486845submit query :2-1-1562763486846submit query :3-1-1562763486846submit query :4-1-1562763486849submit query :5-1-1562763486849submit query :6-1-1562763486859submit query :7-1-1562763486913submit query :8-1-1562763486967submit query :9-1-1562763487021execute query :1-2-phone : 12345678910-1562763487316

1> {"password":"1562763486506","phone":"12345678910","id":"1","username":"venn"}

submit query :10-1-1562763487408submit query :11-1-1562763487408execute query :9-2-phone : 1562661110630-1562763487633

1> {"password":"1562763487017","phone":"1562661110630","id":"9","username":"venn"} # 这里可以看到异步，提交查询的到 11 了，执行查询 的只有 1/9，返回了 1/9(unorderedWait 调用)

submit query :12-1-1562763487634execute query :8-2-phone : 1562661110627-1562763487932

1> {"password":"1562763486963","phone":"1562661110627","id":"8","username":"venn"}

submit query :13-1-1562763487933execute query :7-2-phone : 1562661110624-1562763488228

1> {"password":"1562763486909","phone":"1562661110624","id":"7","username":"venn"}

submit query :14-1-1562763488230execute query :6-2-phone : 1562661110622-1562763488526

1> {"password":"1562763486855","phone":"1562661110622","id":"6","username":"venn"}

submit query :15-1-1562763488527execute query :4-2-phone : 12345678913-1562763488832

1> {"password":"1562763486748","phone":"12345678913","id":"4","username":"venn"}

hbase、redis或其他实现类似

欢迎关注Flink菜鸟公众号，会不定期更新Flink(开发技术)相关的推文