1、为何要 BulkLoad 导入?传统的 HTableOutputFormat 写 HBase 有什么问题?

我们先看下 HBase 的写流程:

hbase多进程入库 hbase入库方式_hbase数据导入

通常 MapReduce 在写HBase时使用的是 TableOutputFormat 方式,在reduce中直接生成put对象写入HBase,该方式在大数据量写入时效率低下(HBase会block写入,频繁进行flush,split,compact等大量IO操作),并对HBase节点的稳定性造成一定的影响(GC时间过长,响应变慢,导致节点超时退出,并引起一系列连锁反应),而HBase支持 bulk load 的入库方式,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,直接在HDFS中生成持久化的HFile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,在大数据量写入时能极大的提高写入效率,并降低对HBase节点的写入压力。
通过使用先生成HFile,然后再BulkLoad到Hbase的方式来替代之前直接调用HTableOutputFormat的方法有如下的好处:
(1)消除了对HBase集群的插入压力
(2)提高了Job的运行速度,降低了Job的执行时间
目前此种方式仅仅适用于只有一个列族的情况,在新版 HBase 中,单列族的限制会消除。

2、bulkload 流程与实践

bulkload 方式需要两个Job配合完成: 
(1)第一个Job还是运行原来业务处理逻辑,处理的结果不直接调用HTableOutputFormat写入到HBase,而是先写入到HDFS上的一个中间目录下(如 middata) 
(2)第二个Job以第一个Job的输出(middata)做为输入,然后将其格式化HBase的底层存储文件HFile 
(3)调用BulkLoad将第二个Job生成的HFile导入到对应的HBase表中

下面给出相应的范例代码:

?

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.client.Put;
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.LoadIncrementalHFiles;
import org.apache.hadoop.hbase.util.Bytes;
import org.apache.hadoop.io.IntWritable;
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.Reducer;
import org.apache.hadoop.mapreduce.lib.input.FileInputFormat;
import org.apache.hadoop.mapreduce.lib.input.TextInputFormat;
import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;
import org.apache.hadoop.mapreduce.lib.output.TextOutputFormat;
import org.apache.hadoop.util.GenericOptionsParser;
 
public class GeneratePutHFileAndBulkLoadToHBase {
 
public static class WordCountMapper extends Mapper<LongWritable, Text, Text, IntWritable>
{
 
private Text wordText=new Text();
private IntWritable one=new IntWritable(1);
@Override
protected void map(LongWritable key, Text value, Context context)
throws IOException, InterruptedException {
// TODO Auto-generated method stub
String line=value.toString();
String[] wordArray=line.split(" ");
for(String word:wordArray)
{
wordText.set(word);
context.write(wordText, one);
}
 
}
}
 
public static class WordCountReducer extends Reducer<Text, IntWritable, Text, IntWritable>
{
 
private IntWritable result=new IntWritable();
protected void reduce(Text key, Iterable<IntWritable> valueList,
Context context)
throws IOException, InterruptedException {
// TODO Auto-generated method stub
int sum=0;
for(IntWritable value:valueList)
{
sum+=value.get();
}
result.set(sum);
context.write(key, result);
}
 
}
 
public static class ConvertWordCountOutToHFileMapper extends Mapper<LongWritable, Text, ImmutableBytesWritable, Put>
{
 
@Override
protected void map(LongWritable key, Text value, Context context)
throws IOException, InterruptedException {
// TODO Auto-generated method stub
String wordCountStr=value.toString();
String[] wordCountArray=wordCountStr.split("\t");
String word=wordCountArray[0];
int count=Integer.valueOf(wordCountArray[1]);
 
//创建HBase中的RowKey
byte[] rowKey=Bytes.toBytes(word);
ImmutableBytesWritable rowKeyWritable=new ImmutableBytesWritable(rowKey);
byte[] family=Bytes.toBytes("cf");
byte[] qualifier=Bytes.toBytes("count");
byte[] hbaseValue=Bytes.toBytes(count);
// Put 用于列簇下的多列提交,若只有一个列,则可以使用 KeyValue 格式
// KeyValue keyValue = new KeyValue(rowKey, family, qualifier, hbaseValue);
Put put=new Put(rowKey);
put.add(family, qualifier, hbaseValue);
context.write(rowKeyWritable, put);
 
}
 
}
 
public static void main(String[] args) throws Exception {
// TODO Auto-generated method stub
Configuration hadoopConfiguration=new Configuration();
String[] dfsArgs = new GenericOptionsParser(hadoopConfiguration, args).getRemainingArgs();
 
//第一个Job就是普通MR,输出到指定的目录
Job job=new Job(hadoopConfiguration, "wordCountJob");
job.setJarByClass(GeneratePutHFileAndBulkLoadToHBase.class);
job.setMapperClass(WordCountMapper.class);
job.setReducerClass(WordCountReducer.class);
job.setOutputKeyClass(Text.class);
job.setOutputValueClass(IntWritable.class);
FileInputFormat.setInputPaths(job, new Path(dfsArgs[0]));
FileOutputFormat.setOutputPath(job, new Path(dfsArgs[1]));
//提交第一个Job
int wordCountJobResult=job.waitForCompletion(true)?0:1;
 
//第二个Job以第一个Job的输出做为输入,只需要编写Mapper类,在Mapper类中对一个job的输出进行分析,并转换为HBase需要的KeyValue的方式。
Job convertWordCountJobOutputToHFileJob=new Job(hadoopConfiguration, "wordCount_bulkload");
 
convertWordCountJobOutputToHFileJob.setJarByClass(GeneratePutHFileAndBulkLoadToHBase.class);
convertWordCountJobOutputToHFileJob.setMapperClass(ConvertWordCountOutToHFileMapper.class);
//ReducerClass 无需指定,框架会自行根据 MapOutputValueClass 来决定是使用 KeyValueSortReducer 还是 PutSortReducer
//convertWordCountJobOutputToHFileJob.setReducerClass(KeyValueSortReducer.class);
convertWordCountJobOutputToHFileJob.setMapOutputKeyClass(ImmutableBytesWritable.class);
convertWordCountJobOutputToHFileJob.setMapOutputValueClass(Put.class);
 
//以第一个Job的输出做为第二个Job的输入
FileInputFormat.addInputPath(convertWordCountJobOutputToHFileJob, new Path(dfsArgs[1]));
FileOutputFormat.setOutputPath(convertWordCountJobOutputToHFileJob, new Path(dfsArgs[2]));
//创建HBase的配置对象
Configuration hbaseConfiguration=HBaseConfiguration.create();
//创建目标表对象
HTable wordCountTable =new HTable(hbaseConfiguration, "word_count");
HFileOutputFormat.configureIncrementalLoad(convertWordCountJobOutputToHFileJob,wordCountTable);
 
//提交第二个job
int convertWordCountJobOutputToHFileJobResult=convertWordCountJobOutputToHFileJob.waitForCompletion(true)?0:1;
 
//当第二个job结束之后,调用BulkLoad方式来将MR结果批量入库
LoadIncrementalHFiles loader = new LoadIncrementalHFiles(hbaseConfiguration);
//第一个参数为第二个Job的输出目录即保存HFile的目录,第二个参数为目标表
loader.doBulkLoad(new Path(dfsArgs[2]), wordCountTable);
 
//最后调用System.exit进行退出
System.exit(convertWordCountJobOutputToHFileJobResult);
 
}
 
}

比如原始的输入数据的目录为:/rawdata/test/wordcount/20131212 

中间结果数据保存的目录为:/middata/test/wordcount/20131212 
最终生成的HFile保存的目录为:/resultdata/test/wordcount/20131212 
运行上面的Job的方式如下: 
hadoop jar test.jar /rawdata/test/wordcount/20131212 /middata/test/wordcount/20131212 /resultdata/test/wordcount/20131212 

3、说明与注意事项:

(1)HFile方式在所有的加载方案里面是最快的,不过有个前提——数据是第一次导入,表是空的。如果表中已经有了数据。HFile再导入到hbase的表中会触发split操作。

(2)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>。
否则报这样的错误:

?

java.lang.IllegalArgumentException: Can't read partitions file
...
Caused by: java.io.IOException: wrong key class: org.apache.hadoop.io.*** is not class org.apache.hadoop.hbase.io.ImmutableBytesWritable

(3)最终输出部分,Value类型是KeyValue 或Put,对应的Sorter分别是KeyValueSortReducer或PutSortReducer,这个 SorterReducer 可以不指定,因为源码中已经做了判断:

?

if (KeyValue.class.equals(job.getMapOutputValueClass())) {
job.setReducerClass(KeyValueSortReducer.class);
} else if (Put.class.equals(job.getMapOutputValueClass())) {
job.setReducerClass(PutSortReducer.class);
} else {
LOG.warn("Unknown map output value type:" + job.getMapOutputValueClass());
}

(4) MR例子中job.setOutputFormatClass(HFileOutputFormat.class); HFileOutputFormat只适合一次对单列族组织成HFile文件,多列簇需要起多个 job,不过新版本的 Hbase 已经解决了这个限制。 

(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。

(6)最后一个 Reduce 没有 setNumReduceTasks 是因为,该设置由框架根据region个数自动配置的。

(7)下边配置部分,注释掉的其实写不写都无所谓,因为看源码就知道configureIncrementalLoad方法已经把固定的配置全配置完了,不固定的部分才需要手动配置。

?

public class HFileOutput {
//job 配置
public static Job configureJob(Configuration conf) throws IOException {
Job job = new Job(configuration, "countUnite1");
job.setJarByClass(HFileOutput.class);
//job.setNumReduceTasks(2);  
//job.setOutputKeyClass(ImmutableBytesWritable.class);
//job.setOutputValueClass(KeyValue.class);
//job.setOutputFormatClass(HFileOutputFormat.class);
 
Scan scan = new Scan();
scan.setCaching(10);
scan.addFamily(INPUT_FAMILY);
TableMapReduceUtil.initTableMapperJob(inputTable, scan,
HFileOutputMapper.class, ImmutableBytesWritable.class, LongWritable.class, job);
//这里如果不定义reducer部分,会自动识别定义成KeyValueSortReducer.class 和PutSortReducer.class
job.setReducerClass(HFileOutputRedcuer.class);
//job.setOutputFormatClass(HFileOutputFormat.class);
HFileOutputFormat.configureIncrementalLoad(job, new HTable(
configuration, outputTable));
HFileOutputFormat.setOutputPath(job, new Path());
//FileOutputFormat.setOutputPath(job, new Path()); //等同上句
return job;
}
 
public static class HFileOutputMapper extends
TableMapper<ImmutableBytesWritable, LongWritable> {
public void map(ImmutableBytesWritable key, Result values,
Context context) throws IOException, InterruptedException {
//mapper逻辑部分
context.write(new ImmutableBytesWritable(Bytes()), LongWritable());
}
}
 
public static class HFileOutputRedcuer extends
Reducer<ImmutableBytesWritable, LongWritable, ImmutableBytesWritable, KeyValue> {
public void reduce(ImmutableBytesWritable key, Iterable<LongWritable> values,
Context context) throws IOException, InterruptedException {
//reducer逻辑部分
KeyValue kv = new KeyValue(row, OUTPUT_FAMILY, tmp[1].getBytes(),
Bytes.toBytes(count));
context.write(key, kv);
}
}
}

4、Refer:

1、Hbase几种数据入库(load)方式比较

2、MapReduce生成HFile入库到HBase及源码分析

http://blog.pureisle.net/archives/1950.html

3、MapReduce生成HFile入库到HBase

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