HBase數據快速導入之ImportTsv&Bulkload

導入數據最快的方式，能夠略過WAL直接生產底層HFile文件

（環境：centos6.五、Hadoop2.6.0、HBase0.98.9）

1.SHELL方式

1.1 ImportTsv直接導入

命令：bin/hbase org.apache.hadoop.hbase.mapreduce.ImportTsv

Usage: importtsv -Dimporttsv.columns=a,b,c <tablename> <inputdir>

測試：

1.1.1在HBase中建立好表

create ‘testImport1’,’cf’

1.1.2準備數據文件sample1.csv，並上傳到HDFS，內容爲：

1,"tom"
2,"sam"
3,"jerry"
4,"marry"
5,"john

1.1.3使用導入命令導入

bin/hbase org.apache.hadoop.hbase.mapreduce.ImportTsv -Dimporttsv.separator="," -Dimporttsv.columns=HBASE_ROW_KEY,cf testImport1 /sample1.csv

1.1.4結果

 

1.2先經過ImportTsv生產HFile文件，再經過completeBulkload導入HBase

1.2.1使用剛纔的源數據並建立新表

create ‘testImport2’,’cf’

1.2.2使用命令生產HFile文件

bin/hbase org.apache.hadoop.hbase.mapreduce.ImportTsv -Dimporttsv.separator="," -Dimporttsv.bulk.output=hfile_tmp -Dimporttsv.columns=HBASE_ROW_KEY,cf testImport2 /sample1.csv

1.2.3在HDFS上的中間結果

1.2.4使用命令將HFile文件導入HBase

hadoop jar lib/hbase-server-0.98.9-hadoop2.jar completebulkload hfile_tmp testImport2

1.2.5結果

注：1.若是出現缺包錯誤提示，則把HBase的jar包包含到hadoop的classpath中；2.運行該命令的本質是一個hdfs的mv操做，並不會啓動MapReduce。

2.API代碼方式

代碼的方式更靈活一點，許多東西能夠自定義。

直接貼代碼吧：

import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.FsShell;
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.HFileOutputFormat2;
import org.apache.hadoop.hbase.mapreduce.LoadIncrementalHFiles;
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.input.TextInputFormat;
import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;
import org.apache.hadoop.util.GenericOptionsParser;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;

import java.io.IOException;

public class BulkLoadJob {
	static Logger logger = LoggerFactory.getLogger(BulkLoadJob.class);

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

		public void map(LongWritable key, Text value, Context context) throws IOException, InterruptedException {

			String[] valueStrSplit = value.toString().split("\t");
			String hkey = valueStrSplit[0];
			String family = valueStrSplit[1].split(":")[0];
			String column = valueStrSplit[1].split(":")[1];
			String hvalue = valueStrSplit[2];
			final byte[] rowKey = Bytes.toBytes(hkey);
			final ImmutableBytesWritable HKey = new ImmutableBytesWritable(rowKey);
			// Put HPut = new Put(rowKey);
			// byte[] cell = Bytes.toBytes(hvalue);
			// HPut.add(Bytes.toBytes(family), Bytes.toBytes(column), cell);
			KeyValue kv = new KeyValue(rowKey, Bytes.toBytes(family), Bytes.toBytes(column), Bytes.toBytes(hvalue));
			context.write(HKey, kv);
		}
	}

	public static void main(String[] args) throws Exception {
		Configuration conf = HBaseConfiguration.create();
		conf.set("hbase.zookeeper.property.clientPort", "2182");
		conf.set("hbase.zookeeper.quorum", "msg801,msg802,msg803");
		conf.set("hbase.master", "msg801:60000");
		String[] dfsArgs = new GenericOptionsParser(conf, args).getRemainingArgs();
		String inputPath = dfsArgs[0];
		System.out.println("source: " + dfsArgs[0]);
		String outputPath = dfsArgs[1];
		System.out.println("dest: " + dfsArgs[1]);
		HTable hTable = null;
		try {
			Job job = Job.getInstance(conf, "Test Import HFile & Bulkload");
			job.setJarByClass(BulkLoadJob.class);
			job.setMapperClass(BulkLoadJob.BulkLoadMap.class);
			job.setMapOutputKeyClass(ImmutableBytesWritable.class);
			job.setMapOutputValueClass(KeyValue.class);
			// speculation
			job.setSpeculativeExecution(false);
			job.setReduceSpeculativeExecution(false);
			// in/out format
			job.setInputFormatClass(TextInputFormat.class);
			job.setOutputFormatClass(HFileOutputFormat2.class);

			FileInputFormat.setInputPaths(job, inputPath);
			FileOutputFormat.setOutputPath(job, new Path(outputPath));

			hTable = new HTable(conf, dfsArgs[2]);
			HFileOutputFormat2.configureIncrementalLoad(job, hTable);

			if (job.waitForCompletion(true)) {
				FsShell shell = new FsShell(conf);
				try {
					shell.run(new String[] { "-chmod", "-R", "777", dfsArgs[1] });
				} catch (Exception e) {
					logger.error("Couldnt change the file permissions ", e);
					throw new IOException(e);
				}
				// 加載到hbase表
				LoadIncrementalHFiles loader = new LoadIncrementalHFiles(conf);
				// 兩種方式均可以
				// 方式一
				String[] loadArgs = { outputPath, dfsArgs[2] };
				loader.run(loadArgs);
				// 方式二
				// loader.doBulkLoad(new Path(outputPath), hTable);
			} else {
				logger.error("loading failed.");
				System.exit(1);
			}

		} catch (IllegalArgumentException e) {
			e.printStackTrace();
		} finally {
			if (hTable != null) {
				hTable.close();
			}
		}
	}
}

　　

2.1建立新表

create ‘testImport3’,’fm1’,’fm2’

2.2建立sample2.csv，並上傳到HDFS，內容爲：
key1    fm1:col1        value1
key1    fm1:col2        value2
key1    fm2:col1        value3
key4    fm1:col1        value4

使用命令：

hadoop jar BulkLoadJob.jar hdfs://msg/sample2,csv hdfs://msg/HFileOut testImport3

注：1.mapper中使用KeyValue和Put均可以；2.注意jar包的classpath；3.若是Hadoop是HA，則須要使用HA的名字，好比咱們的active namenode名稱爲msg801，可是HA的nameservice爲msg，則HDFS的路徑必須使用hdfs://msg而不能使用hdfs://msg801:9000（WHY？）。

具體報錯爲：

IllegalArgumentException: Wrong FS: hdfs://msg801:9000/HFileOut/fm2/bbab9d883a574d518cdcb304d1e681e9, expected: hdfs://msg