使用BulkLoad從HDFS批量導(dǎo)入數(shù)據(jù)到HBase

在向Hbase中寫入數(shù)據(jù)時(shí)，常見的寫入方法有使用HBase API，Mapreduce批量導(dǎo)入數(shù)據(jù)，使用這些方式帶入數(shù)據(jù)時(shí)，一條數(shù)據(jù)寫入到HBase數(shù)據(jù)庫中的大致流程如圖。

數(shù)據(jù)發(fā)出后首先寫入到雨鞋日志W(wǎng)Al中，寫入到預(yù)寫日志中之后，隨后寫入到內(nèi)存MemStore中，最后在Flush到Hfile中。這樣寫數(shù)據(jù)的方式不會(huì)導(dǎo)致數(shù)據(jù)的丟失，并且道正數(shù)據(jù)的有序性，但是當(dāng)遇到大量的數(shù)據(jù)寫入時(shí)，寫入的速度就難以保證。所以，介紹一種性能更高的寫入方式BulkLoad。

使用BulkLoad批量寫入數(shù)據(jù)主要分為兩部分：
一、使用HFileOutputFormat2通過自己編寫的MapReduce作業(yè)將HFile寫入到HDFS目錄，由于寫入到HBase中的數(shù)據(jù)是按照順序排序的，HFileOutputFormat2中的configureIncrementalLoad()可以完成所需的配置。
二、將Hfile從HDFS移動(dòng)到HBase表中，大致過程如圖

實(shí)例代碼pom依賴：

<dependency>
            <groupId>org.apache.hbase</groupId>
            <artifactId>hbase-server</artifactId>
            <version>1.4.0</version>
        </dependency>

        <dependency>
            <groupId>org.apache.hadoop</groupId>
            <artifactId>hadoop-client</artifactId>
            <version>2.6.4</version>
        </dependency>

        <dependency>
            <groupId>org.apache.hbase</groupId>
            <artifactId>hbase-client</artifactId>
            <version>0.99.2</version>
        </dependency>

package com.yangshou;

import org.apache.hadoop.hbase.client.Put;
import org.apache.hadoop.hbase.io.ImmutableBytesWritable;
import org.apache.hadoop.io.LongWritable;
import org.apache.hadoop.io.Text;
import org.apache.hadoop.mapreduce.Mapper;

import java.io.IOException;

public class BulkLoadMapper extends Mapper<LongWritable, Text, ImmutableBytesWritable, Put> {
    @Override
    protected void map(LongWritable key, Text value, Context context) throws IOException, InterruptedException {
        //讀取文件中的每一條數(shù)據(jù)，以序號(hào)作為行鍵
        String line = value.toString();
        //將數(shù)據(jù)進(jìn)行切分
        //切分后數(shù)組中的元素分別為：序號(hào)，用戶id,商品id,用戶行為，商品分類，時(shí)間，地址
        String[] str = line.split(" ");
        String id = str[0];
        String user_id = str[1];
        String item_id = str[2];
        String behavior = str[3];
        String item_type = str[4];
        String time = str[5];
        String address = "156";
        //拼接rowkey和put
        ImmutableBytesWritable rowkry = new ImmutableBytesWritable(id.getBytes());
        Put put = new Put(id.getBytes());
        put.add("info".getBytes(),"user_id".getBytes(),user_id.getBytes());
        put.add("info".getBytes(),"item_id".getBytes(),item_id.getBytes());
        put.add("info".getBytes(),"behavior".getBytes(),behavior.getBytes());
        put.add("info".getBytes(),"item_type".getBytes(),item_type.getBytes());
        put.add("info".getBytes(),"time".getBytes(),time.getBytes());
        put.add("info".getBytes(),"address".getBytes(),address.getBytes());
        //將數(shù)據(jù)寫出
        context.write(rowkry,put);
    }
}

package com.yangshou;

import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.hbase.HBaseConfiguration;
import org.apache.hadoop.hbase.TableName;
import org.apache.hadoop.hbase.client.*;
import org.apache.hadoop.hbase.io.ImmutableBytesWritable;
import org.apache.hadoop.hbase.mapreduce.HFileOutputFormat2;
import org.apache.hadoop.hbase.mapreduce.LoadIncrementalHFiles;
import org.apache.hadoop.mapreduce.Job;
import org.apache.hadoop.mapreduce.lib.input.FileInputFormat;
import org.apache.hadoop.mapreduce.lib.input.TextInputFormat;
import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;

public class BulkLoadDriver  {
    public static void main(String[] args) throws Exception {
        //獲取Hbase配置
        Configuration conf = HBaseConfiguration.create();
        Connection conn = ConnectionFactory.createConnection(conf);
        Table table = conn.getTable(TableName.valueOf("BulkLoadDemo"));
        Admin admin = conn.getAdmin();

        //設(shè)置job
        Job job = Job.getInstance(conf,"BulkLoad");
        job.setJarByClass(BulkLoadDriver.class);
        job.setMapperClass(BulkLoadMapper.class);
        job.setMapOutputKeyClass(ImmutableBytesWritable.class);
        job.setMapOutputValueClass(Put.class);

        //設(shè)置文件的輸入輸出路徑
        job.setInputFormatClass(TextInputFormat.class);
        job.setOutputFormatClass(HFileOutputFormat2.class);
        FileInputFormat.setInputPaths(job,new Path("hdfs://hadoopalone:9000/tmp/000000_0"));
        FileOutputFormat.setOutputPath(job,new Path("hdfs://hadoopalone:9000/demo1"));

        //將數(shù)據(jù)加載到Hbase表中
        HFileOutputFormat2.configureIncrementalLoad(job,table,conn.getRegionLocator(TableName.valueOf("BulkLoadDemo")));
        if(job.waitForCompletion(true)){
            LoadIncrementalHFiles load = new LoadIncrementalHFiles(conf);
            load.doBulkLoad(new Path("hdfs://hadoopalone:9000/demo1"),admin,table,conn.getRegionLocator(TableName.valueOf("BulkLoadDemo")));

        }

    }
}

實(shí)例數(shù)據(jù)

44979   100640791   134060896   1   5271    2014-12-09  天津市
44980   100640791   96243605    1   13729   2014-12-02  新疆

在Hbase shell 中創(chuàng)建表

create 'BulkLoadDemo','info'

打包后執(zhí)行
```hadoop jar BulkLoadDemo-1.0-SNAPSHOT.jar com.yangshou.BulkLoadDriver

注意：在執(zhí)行hadoop jar之前應(yīng)該先將Hbase中的相關(guān)包加載過來

export HADOOP_CLASSPATH=$HBASE_HOME/lib/*