0. 前言
Flink被誉为第四代大数据计算引擎组件，即可以用作基于离线分布式计算，也可以应用于实时计算。Flink的核心是转化为流进行计算。Flink三个核心：Source,Transformation,Sink。其中Source即为Flink计算的数据源，Transformation即为进行分布式流式计算的算子，也是计算的核心，Sink即为计算后的数据输出端。Flink Source原生支持包括Kafka,ES,RabbitMQ等一些通用的消息队列组件或基于文本的高性能非关系型数据库。而Flink Sink写原生也只支持类似Redis，Kafka,ES,RabbitMQ等一些通用的消息队列组件或基于文本的高性能非关系型数据库。而对于写入关系型数据库或Flink不支持的组件中，需要借助RichSourceFunction去实现，但这部分性能是比原生的差些，虽然Flink不建议这么做，但在大数据处理过程中，由于业务或技术架构的复杂性，有些特定的场景还是需要这样做，本篇博客就是介绍如何通过Flink RichSourceFunction来写关系型数据库，这里以写mysql为例。
1. 引入依赖的jar包
flink基础包
flink-jdbc包
mysql-jdbc包
2. 继承RichSourceFunction包将jdbc封装读mysql
package com.run;
import java.sql.DriverManager;
import java.sql.ResultSet;
import org.apache.flink.api.java.tuple.Tuple10;
import org.apache.flink.configuration.Configuration;
import org.apache.flink.streaming.api.functions.source.RichSourceFunction;
import com.mysql.jdbc.Connection;
import com.mysql.jdbc.PreparedStatement;
public class Flink2JdbcReader extends
RichSourceFunction<Tuple10<String, String, String, String, String, String, String, String, String, String>> {
private static final long serialVersionUID = 3334654984018091675L;
private Connection connect = null;
private PreparedStatement ps = null;
/*
* (non-Javadoc)
*
* @see org.apache.flink.api.common.functions.AbstractRichFunction#open(org.
* apache.flink.configuration.Configuration) to use open database connect
*/
@Override
public void open(Configuration parameters) throws Exception {
super.open(parameters);
Class.forName("com.mysql.jdbc.Driver");
connect = (Connection) DriverManager.getConnection("jdbc:mysql://192.168.21.11:3306", "root", "flink");
ps = (PreparedStatement) connect
.prepareStatement("select col1,col2,col3,col4,col5,col6,col7,col8,col9,col10 from flink.test_tb");
}
/*
* (non-Javadoc)
*
* @see
* org.apache.flink.streaming.api.functions.source.SourceFunction#run(org.
* apache.flink.streaming.api.functions.source.SourceFunction.SourceContext)
* to use excuted sql and return result
*/
@Override
public void run(
SourceContext<Tuple10<String, String, String, String, String, String, String, String, String, String>> collect)
throws Exception {
ResultSet resultSet = ps.executeQuery();
while (resultSet.next()) {
Tuple10<String, String, String, String, String, String, String, String, String, String> tuple = new Tuple10<String, String, String, String, String, String, String, String, String, String>();
tuple.setFields(resultSet.getString(1), resultSet.getString(2), resultSet.getString(3),
resultSet.getString(4), resultSet.getString(5), resultSet.getString(6), resultSet.getString(7),
resultSet.getString(8), resultSet.getString(9), resultSet.getString(10));
collect.collect(tuple);
}
}
/*
* (non-Javadoc)
*
* @see
* org.apache.flink.streaming.api.functions.source.SourceFunction#cancel()
* colse database connect
*/
@Override
public void cancel() {
try {
super.close();
if (connect != null) {
connect.close();
}
if (ps != null) {
ps.close();
}
} catch (Exception e) {
e.printStackTrace();
}
}
}
3. 继承RichSourceFunction包将jdbc封装写mysql
package com.run;
import java.sql.DriverManager;
import org.apache.flink.api.java.tuple.Tuple10;
import org.apache.flink.configuration.Configuration;
import org.apache.flink.streaming.api.functions.sink.RichSinkFunction;
import com.mysql.jdbc.Connection;
import com.mysql.jdbc.PreparedStatement;
public class Flink2JdbcWriter extends
RichSinkFunction<Tuple10<String, String, String, String, String, String, String, String, String, String>> {
private static final long serialVersionUID = -8930276689109741501L;
private Connection connect = null;
private PreparedStatement ps = null;
/*
* (non-Javadoc)
*
* @see org.apache.flink.api.common.functions.AbstractRichFunction#open(org.
* apache.flink.configuration.Configuration) get database connect
*/
@Override
public void open(Configuration parameters) throws Exception {
super.open(parameters);
Class.forName("com.mysql.jdbc.Driver");
connect = (Connection) DriverManager.getConnection("jdbc:mysql://192.168.21.11:3306", "root", "flink");
ps = (PreparedStatement) connect.prepareStatement("insert into flink.test_tb1 values (?,?,?,?,?,?,?,?,?,?)");
}
/*
* (non-Javadoc)
*
* @see
* org.apache.flink.streaming.api.functions.sink.SinkFunction#invoke(java.
* lang.Object,
* org.apache.flink.streaming.api.functions.sink.SinkFunction.Context) read
* data from flink DataSet to database
*/
@Override
public void invoke(Tuple10<String, String, String, String, String, String, String, String, String, String> value,
Context context) throws Exception {
ps.setString(1, value.f0);
ps.setString(2, value.f1);
ps.setString(3, value.f2);
ps.setString(4, value.f3);
ps.setString(5, value.f4);
ps.setString(6, value.f5);
ps.setString(7, value.f6);
ps.setString(8, value.f7);
ps.setString(9, value.f8);
ps.setString(10, value.f9);
ps.executeUpdate();
}
/*
* (non-Javadoc)
*
* @see org.apache.flink.api.common.functions.AbstractRichFunction#close()
* close database connect
*/
@Override
public void close() throws Exception {
try {
super.close();
if (connect != null) {
connect.close();
}
if (ps != null) {
ps.close();
}
} catch (Exception e) {
e.printStackTrace();
}
}
}
4. 代码解释
对于Flink2JdbcReader的读
里面有三个方法open，run，cancel，其中open方法是建立与关系型数据库的链接，这里其实就是普通的jdbc链接及mysql的地址，端口，库等信息。run方法是读取mysql数据转化为Flink独有的Tuple集合类型，可以根据代码看出其中的规律和Tuple8，Tuple9，Tuple10代表什么含义。cancel就很简单了关闭数据库连接
对于Flink2JdbcWriter的写
里面有三个方法open，invoke，close，其中open方法是建立与关系型数据库的链接，这里其实就是普通的jdbc链接及mysql的地址，端口，库等信息。invoke方法是将flink的数据类型插入到mysql，这里的写法与在web程序中写jdbc插入数据不太一样，因为flink独有的Tuple，可以根据代码看出其中的规律和Tuple8，Tuple9，Tuple10代表什么含义。close关闭数据库连接
5. 测试:读mysql数据并继续写入mysql
package com.run;
import java.util.Date;
import org.apache.flink.api.java.tuple.Tuple10;
import org.apache.flink.streaming.api.datastream.DataStream;
import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
public class FlinkReadDbWriterDb {
public static void main(String[] args) throws Exception {。
final StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
DataStream<Tuple10<String, String, String, String, String, String, String, String, String, String>> dataStream = env
.addSource(new Flink2JdbcReader());
// tranfomat
dataStream.addSink(new Flink2JdbcWriter());
env.execute("Flink cost DB data to write Database") ;
}
}
6. 总结
从测试代码中可以很清晰的看出Flink的逻辑:Source->Transformation->Sink,可以在addSource到addSink之间加入我们的业务逻辑算子。同时这里必须注意env.execute("Flink cost DB data to write Database");这个必须有而且必须要放到结尾，否则整个代码是不会执行的，至于为什么在后续的博客会讲
