MapReduce生成HFile入库到HBase

原文：http://shitouer.cn/2013/02/hbase-hfile-bulk-load/

 

一、这种方式有很多的优点：

1. 如果我们一次性入库hbase巨量数据，处理速度慢不说，还特别占用Region资源， 一个比较高效便捷的方法就是使用 “Bulk Loading”方法，即HBase提供的HFileOutputFormat类。

2. 它是利用hbase的数据信息按照特定格式存储在hdfs内这一原理，直接生成这种hdfs内存储的数据格式文件，然后上传至合适位置，即完成巨量数据快速入库的办法。配合mapreduce完成，高效便捷，而且不占用region资源，增添负载。

二、这种方式也有很大的限制：

1. 仅适合初次数据导入，即表内数据为空，或者每次入库表内都无数据的情况。

2. HBase集群与Hadoop集群为同一集群，即HBase所基于的HDFS为生成HFile的MR的集群(额，咋表述~~~)

三、接下来一个demo，简单介绍整个过程。

1. 生成HFile部分

package zl.hbase.mr;

import java.io.IOException;

import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.hbase.KeyValue;
import org.apache.hadoop.hbase.io.ImmutableBytesWritable;
import org.apache.hadoop.hbase.mapreduce.HFileOutputFormat;
import org.apache.hadoop.hbase.mapreduce.KeyValueSortReducer;
import org.apache.hadoop.hbase.mapreduce.SimpleTotalOrderPartitioner;
import org.apache.hadoop.hbase.util.Bytes;
import org.apache.hadoop.io.LongWritable;
import org.apache.hadoop.io.Text;
import org.apache.hadoop.mapreduce.Job;
import org.apache.hadoop.mapreduce.Mapper;
import org.apache.hadoop.mapreduce.lib.input.FileInputFormat;
import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;
import org.apache.hadoop.util.GenericOptionsParser;

import zl.hbase.util.ConnectionUtil;

public class HFileGenerator {

    public static class HFileMapper extends
            Mapper<LongWritable, Text, ImmutableBytesWritable, KeyValue> {
        @Override
        protected void map(LongWritable key, Text value, Context context)
                throws IOException, InterruptedException {
            String line = value.toString();
            String[] items = line.split( "," , - 1 );
            ImmutableBytesWritable rowkey = new ImmutableBytesWritable(
                    items[ 0 ].getBytes());

            KeyValue kv = new KeyValue(Bytes.toBytes(items[ 0 ]),
                    Bytes.toBytes(items[ 1 ]), Bytes.toBytes(items[ 2 ]),
                    System.currentTimeMillis(), Bytes.toBytes(items[ 3 ]));
            if ( null != kv) {
                context.write(rowkey, kv);
            }
        }
    }

    public static void main(String[] args) throws IOException,
            InterruptedException, ClassNotFoundException {
        Configuration conf = new Configuration();
        String[] dfsArgs = new GenericOptionsParser(conf, args)
                .getRemainingArgs();

        Job job = new Job(conf, "HFile bulk load test" );
        job.setJarByClass(HFileGenerator. class );

        job.setMapperClass(HFileMapper. class );
        job.setReducerClass(KeyValueSortReducer. class );

        job.setMapOutputKeyClass(ImmutableBytesWritable. class );
        job.setMapOutputValueClass(Text. class );

        job.setPartitionerClass(SimpleTotalOrderPartitioner. class );

        FileInputFormat.addInputPath(job, new Path(dfsArgs[ 0 ]));
        FileOutputFormat.setOutputPath(job, new Path(dfsArgs[ 1 ]));

        HFileOutputFormat.configureIncrementalLoad(job,
                ConnectionUtil.getTable());
        System.exit(job.waitForCompletion( true ) ? 0 : 1 );
    }
}

生成HFile程序说明：

①. 最终输出结果，无论是map还是reduce，输出部分key和value的类型必须是： < ImmutableBytesWritable, KeyValue>或者< ImmutableBytesWritable, Put>。

②. 最终输出部分，Value类型是KeyValue 或Put，对应的Sorter分别是KeyValueSortReducer或PutSortReducer。

③. MR例子中job.setOutputFormatClass(HFileOutputFormat.class); HFileOutputFormat只适合一次对单列族组织成HFile文件。

④. MR例子中HFileOutputFormat.configureIncrementalLoad(job, table);自动对job进行配置。SimpleTotalOrderPartitioner是需要先对key进行整体排序，然后划分到每个reduce中，保证每一个reducer中的的key最小最大值区间范围，是不会有交集的。因为入库到HBase的时候，作为一个整体的Region，key是绝对有序的。

⑤. MR例子中最后生成HFile存储在HDFS上，输出路径下的子目录是各个列族。如果对HFile进行入库HBase，相当于move HFile到HBase的Region中，HFile子目录的列族内容没有了。

2. HFile入库到HBase

 

package zl.hbase.bulkload;

import org.apache.hadoop.fs.Path;
import org.apache.hadoop.hbase.mapreduce.LoadIncrementalHFiles;
import org.apache.hadoop.util.GenericOptionsParser;

import zl.hbase.util.ConnectionUtil;

public class HFileLoader {

    public static void main(String[] args) throws Exception {
        String[] dfsArgs = new GenericOptionsParser(
                ConnectionUtil.getConfiguration(), args).getRemainingArgs();
        LoadIncrementalHFiles loader = new LoadIncrementalHFiles(
                ConnectionUtil.getConfiguration());
        loader.doBulkLoad( new Path(dfsArgs[ 0 ]), ConnectionUtil.getTable());
    }

}

通过HBase中 LoadIncrementalHFiles的doBulkLoad方法，对生成的HFile文件入库

 

 

我修改了一下如下：

import java.io.IOException;

import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.hbase.HBaseConfiguration;
import org.apache.hadoop.hbase.KeyValue;
import org.apache.hadoop.hbase.client.HTable;
import org.apache.hadoop.hbase.io.ImmutableBytesWritable;
import org.apache.hadoop.hbase.mapreduce.HFileOutputFormat;
import org.apache.hadoop.hbase.mapreduce.SimpleTotalOrderPartitioner;
import org.apache.hadoop.hbase.util.Bytes;
import org.apache.hadoop.io.LongWritable;
import org.apache.hadoop.io.Text;
import org.apache.hadoop.mapreduce.Job;
import org.apache.hadoop.mapreduce.Mapper;
import org.apache.hadoop.mapreduce.lib.input.FileInputFormat;
import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;
import org.apache.hadoop.util.GenericOptionsParser;

public class HFileGenerator {

        public static class HFileMapper extends
                        Mapper<LongWritable, Text, ImmutableBytesWritable, KeyValue> {

                ImmutableBytesWritable tableKey = new ImmutableBytesWritable();

                @Override
                protected void map(LongWritable key, Text value, Context context)
                                throws IOException, InterruptedException {

                        String line = value.toString();
                        String[] items = line.split(",", -1);
                        tableKey.set(Bytes.toBytes(items[0]));
                        KeyValue kv = new KeyValue(Bytes.toBytes(items[0]),
                                        Bytes.toBytes(items[1]), Bytes.toBytes(items[2]),
                                        System.currentTimeMillis(), Bytes.toBytes(items[3]));

                        if (kv != null) {
                                context.write(tableKey, kv);
                        }

                }

        }

        /**
         * * @param args * @throws IOException
         * */
        public static void main(String[] args) throws Exception {

                Configuration conf = new Configuration();
                String[] otherArgs = new GenericOptionsParser(conf, args)
                                .getRemainingArgs();
                if (otherArgs.length != 2) {
                        System.err.println("Usage: " + HFileGenerator.class.getName()
                                        + " <in> <out>");
                        System.exit(2);
                }
                Job job = new Job(conf, "HFile bulk load test");
                job.setJarByClass(HFileGenerator.class);
                job.setMapperClass(HFileMapper.class);
//              job.setReducerClass(KeyValueSortReducer.class);
//              job.setOutputKeyClass(ImmutableBytesWritable.class);
//              job.setOutputValueClass(Text.class);
//              job.setPartitionerClass(SimpleTotalOrderPartitioner.class);
                FileInputFormat.addInputPath(job, new Path(otherArgs[0]));
                FileOutputFormat.setOutputPath(job, new Path(otherArgs[1]));

                Configuration hbaseconfig = null;
                HTable table;
                hbaseconfig = HBaseConfiguration.create();
                table = new HTable(hbaseconfig, "member3");
                HFileOutputFormat.configureIncrementalLoad(job, table);
                job.setPartitionerClass(SimpleTotalOrderPartitioner.class);
                System.exit(job.waitForCompletion(true) ? 0 : 1);

        }
}

// job.setReducerClass(KeyValueSortReducer.class);// job.setOutputKeyClass(ImmutableBytesWritable.class);// job.setOutputValueClass(Text.class);// job.setPartitionerClass(SimpleTotalOrderPartitioner.class);

这几句可以不用写，因为在HFileOutputFormat.configureIncrementalLoad(job, table);会设置它们。

 

2.HFileLoader的修改

import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.hbase.HBaseConfiguration;
import org.apache.hadoop.hbase.client.HTable;
import org.apache.hadoop.hbase.mapreduce.LoadIncrementalHFiles;

public class HFileLoader {

        public static void main(String[] args) throws Exception {
                Configuration hbaseconfig = null;
                HTable table;
                hbaseconfig = HBaseConfiguration.create();
                table = new HTable(hbaseconfig, "member3");

                LoadIncrementalHFiles lf = new LoadIncrementalHFiles(hbaseconfig);
                lf.doBulkLoad(new Path("hdfs://master24:9000/user/hadoop/hbasemapred/out"), table);

        }
}

Path用前一步mapred的输出目录,写全路径。