Hadoop中DBInputFormat和DBOutputFormat使用

一、背景

为了方便MapReduce直接访问关系型数据库（Mysql,Oracle），Hadoop提供了DBInputFormat和DBOutputFormat两个类。通过DBInputFormat类把数据库表数据读入到HDFS，根据DBOutputFormat类把MapReduce产生的结果集导入到数据库表中。

推荐阅读：

Hadoop 中利用 MapReduce 读写 MySQL 数据 http://www.linuxidc.com/Linux/2013-07/88117.htm

二、技术细节

1、DBInputFormat（Mysql为例），先创建表:

CREATE TABLE studentinfo (

  id INTEGER NOT NULL PRIMARY KEY,

  name VARCHAR(32) NOT NULL);

2、DBInputFormat用法如下：

public class DBInput {
  // DROP TABLE IF EXISTS `hadoop`.`studentinfo`;
  // CREATE TABLE studentinfo (
  // id INTEGER NOT NULL PRIMARY KEY,
  // name VARCHAR(32) NOT NULL);

  public static class StudentinfoRecord implements Writable, DBWritable {
    int id;
    String name;
    public StudentinfoRecord() {

    }
    public void readFields(DataInput in) throws IOException {
        this.id = in.readInt();
        this.name = Text.readString(in);
    }
    public void write(DataOutput out) throws IOException {
        out.writeInt(this.id);
        Text.writeString(out, this.name);
    }
    public void readFields(ResultSet result) throws SQLException {
        this.id = result.getInt(1);
        this.name = result.getString(2);
    }
    public void write(PreparedStatement stmt) throws SQLException {
        stmt.setInt(1, this.id);
        stmt.setString(2, this.name);
    }
    public String toString() {
        return new String(this.id + " " + this.name);
    }
  }
  public class DBInputMapper extends MapReduceBase implements
        Mapper<LongWritable, StudentinfoRecord, LongWritable, Text> {
    public void map(LongWritable key, StudentinfoRecord value,
          OutputCollector<LongWritable, Text> collector, Reporter reporter)
          throws IOException {
        collector.collect(new LongWritable(value.id), new Text(value
            .toString()));
    }
  }
  public static void main(String[] args) throws IOException {
    JobConf conf = new JobConf(DBInput.class);
    DistributedCache.addFileToClassPath(new Path(
          "/lib/mysql-connector-java-5.1.0-bin.jar"), conf);
   
    conf.setMapperClass(DBInputMapper.class);
    conf.setReducerClass(IdentityReducer.class);

    conf.setMapOutputKeyClass(LongWritable.class);
    conf.setMapOutputValueClass(Text.class);
    conf.setOutputKeyClass(LongWritable.class);
    conf.setOutputValueClass(Text.class);
   
    conf.setInputFormat(DBInputFormat.class);
    FileOutputFormat.setOutputPath(conf, new Path("/hua01"));
    DBConfiguration.configureDB(conf, "com.mysql.jdbc.Driver",
          "jdbc:mysql://192.168.3.244:3306/hadoop", "hua", "hadoop");
    String[] fields = { "id", "name" };
    DBInputFormat.setInput(conf, StudentinfoRecord.class, "studentinfo",
 null, "id", fields);

    JobClient.runJob(conf);
  }
}

a)StudnetinfoRecord类的变量为表字段，实现Writable和DBWritable两个接口。

实现Writable的方法：

 public void readFields(DataInput in) throws IOException {
        this.id = in.readInt();
        this.name = Text.readString(in);
    }
    public void write(DataOutput out) throws IOException {
        out.writeInt(this.id);
        Text.writeString(out, this.name);
    }

实现DBWritable的方法：

public void readFields(ResultSet result) throws SQLException {
        this.id = result.getInt(1);
        this.name = result.getString(2);
    }
    public void write(PreparedStatement stmt) throws SQLException {
        stmt.setInt(1, this.id);
        stmt.setString(2, this.name);
    }

b)读入Mapper的value类型是StudnetinfoRecord。

c)配置如何连入数据库，读出表studentinfo数据。

    DBConfiguration.configureDB(conf, "com.mysql.jdbc.Driver",

             "jdbc:mysql://192.168.3.244:3306/hadoop", "hua", "hadoop");
    String[] fields = { "id", "name" };
    DBInputFormat.setInput(conf, StudentinfoRecord.class, "studentinfo",  null, "id", fields);

3、DBOutputFormat用法如下：

public class DBOutput {

  public static class StudentinfoRecord implements Writable,  DBWritable {
    int id;
    String name;
    public StudentinfoRecord() {

    }
    public void readFields(DataInput in) throws IOException {
        this.id = in.readInt();
        this.name = Text.readString(in);
    }
    public void write(DataOutput out) throws IOException {
        out.writeInt(this.id);
        Text.writeString(out, this.name);
    }
    public void readFields(ResultSet result) throws SQLException {
        this.id = result.getInt(1);
        this.name = result.getString(2);
    }
    public void write(PreparedStatement stmt) throws SQLException {
        stmt.setInt(1, this.id);
        stmt.setString(2, this.name);
    }
    public String toString() {
        return new String(this.id + " " + this.name);
    }
  }
 
  public static class MyReducer extends MapReduceBase implements
        Reducer<LongWritable, Text, StudentinfoRecord, Text> {
    public void reduce(LongWritable key, Iterator<Text> values,
          OutputCollector<StudentinfoRecord, Text> output, Reporter  reporter)
          throws IOException {
        String[] splits = values.next().toString().split("/t");
        StudentinfoRecord r = new StudentinfoRecord();
        r.id = Integer.parseInt(splits[0]);
        r.name = splits[1];
        output.collect(r, new Text(r.name));
    }
  }

  public static void main(String[] args) throws IOException {
    JobConf conf = new JobConf(DBOutput.class);
    conf.setInputFormat(TextInputFormat.class);
    conf.setOutputFormat(DBOutputFormat.class);

    FileInputFormat.setInputPaths(conf, new Path("/hua/hua.bcp"));
    DBConfiguration.configureDB(conf, "com.mysql.jdbc.Driver",
          "jdbc:mysql://192.168.3.244:3306/Hadoop", "hua", "hadoop");
    DBOutputFormat.setOutput(conf, "studentinfo", "id", "name");

  conf.setMapperClass(org.apache.hadoop.mapred.lib.IdentityMapper.class);
    conf.setReducerClass(MyReducer.class);

    JobClient.runJob(conf);
  }

}

a)StudnetinfoRecord类的变量为表字段，实现Writable和DBWritable两个接口，同.DBInputFormat的StudnetinfoRecord类。

b)输出Reducer的key/value类型是StudnetinfoRecord。

c)配置如何连入数据库，输出结果到表studentinfo。

DBConfiguration.configureDB(conf, "com.mysql.jdbc.Driver",
          "jdbc:mysql://192.168.3.244:3306/hadoop", "hua", "hadoop");
    DBOutputFormat.setOutput(conf, "studentinfo", "id", "name");

三、总结

运行MapReduce时候报错：java.io.IOException: com.mysql.jdbc.Driver，一般是由于程序找不到mysql驱动包。解决方法是让每个

tasktracker运行MapReduce程序时都可以找到该驱动包。

添加包有两种方式：

1.在每个节点下的${HADOOP_HOME}/lib下添加该包。重启集群，一般是比较原始的方法。

2.a)把包传到集群上： hadoop fs -put mysql-connector-java-5.1.0- bin.jar /lib

b)在mr程序提交job前，添加语句：istributedCache.addFileToClassPath(new Path("/lib/mysql- connector-java- 5.1.0-bin.jar"), conf);