Hadoop实例WordCount程序一步一步运行

虽说现在用Eclipse下开发Hadoop程序很方便了，但是命令行方式对于小程序开发验证很方便。这是初学hadoop时的笔记，记录下来以备查。

　　1. 经典的WordCound程序（WordCount.java），可参见 hadoop0.18文档

import java.io.IOException;
import java.util.ArrayList;
import java.util.Iterator;
import java.util.List;
import java.util.StringTokenizer;

import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.conf.Configured;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.io.IntWritable;
import org.apache.hadoop.io.LongWritable;
import org.apache.hadoop.io.Text;
import org.apache.hadoop.mapred.FileInputFormat;
import org.apache.hadoop.mapred.FileOutputFormat;
import org.apache.hadoop.mapred.JobClient;
import org.apache.hadoop.mapred.JobConf;
import org.apache.hadoop.mapred.MapReduceBase;
import org.apache.hadoop.mapred.Mapper;
import org.apache.hadoop.mapred.OutputCollector;
import org.apache.hadoop.mapred.Reducer;
import org.apache.hadoop.mapred.Reporter;
import org.apache.hadoop.util.Tool;
import org.apache.hadoop.util.ToolRunner;

public class WordCount extends Configured implements Tool {

public static class MapClass extends MapReduceBase implements
Mapper<LongWritable, Text, Text, IntWritable> {

private final static IntWritable one = new IntWritable(1);
private Text word = new Text();

public void map(LongWritable key, Text value,
OutputCollector<Text, IntWritable> output, Reporter reporter)
throws IOException {
String line = value.toString();
StringTokenizer itr = new StringTokenizer(line);
while (itr.hasMoreTokens()) {
word.set(itr.nextToken());
output.collect(word, one);
}
}
}

/**
* A reducer class that just emits the sum of the input values.
*/
public static class Reduce extends MapReduceBase implements
Reducer<Text, IntWritable, Text, IntWritable> {

public void reduce(Text key, Iterator<IntWritable> values,
OutputCollector<Text, IntWritable> output, Reporter reporter)
throws IOException {
int sum = 0;
while (values.hasNext()) {
sum += values.next().get();
}
output.collect(key, new IntWritable(sum));
}
}

static int printUsage() {
System.out.println("wordcount [-m <maps>] [-r <reduces>] <input> <output>");
ToolRunner.printGenericCommandUsage(System.out);
return -1;
}

/**
* The main driver for word count map/reduce program. Invoke this method to
* submit the map/reduce job.
*
* @throws IOException
* When there is communication problems with the job tracker.
*/
public int run(String[] args) throws Exception {
JobConf conf = new JobConf(getConf(), WordCount.class);
conf.setJobName("wordcount");

// the keys are words (strings)
conf.setOutputKeyClass(Text.class);
// the values are counts (ints)
conf.setOutputValueClass(IntWritable.class);

conf.setMapperClass(MapClass.class);
conf.setCombinerClass(Reduce.class);
conf.setReducerClass(Reduce.class);

List<String> other_args = new ArrayList<String>();
for (int i = 0; i < args.length; ++i) {
try {
if ("-m".equals(args[i])) {
conf.setNumMapTasks(Integer.parseInt(args[++i]));
} else if ("-r".equals(args[i])) {
conf.setNumReduceTasks(Integer.parseInt(args[++i]));
} else {
other_args.add(args[i]);
}
} catch (NumberFormatException except) {
System.out.println("ERROR: Integer expected instead of "
+ args[i]);
return printUsage();
} catch (ArrayIndexOutOfBoundsException except) {
System.out.println("ERROR: Required parameter missing from "
+ args[i - 1]);
return printUsage();
}
}

// Make sure there are exactly 2 parameters left.
if (other_args.size() != 2) {
System.out.println("ERROR: Wrong number of parameters: "
+ other_args.size() + " instead of 2.");
return printUsage();
}
FileInputFormat.setInputPaths(conf, other_args.get(0));
FileOutputFormat.setOutputPath(conf, new Path(other_args.get(1)));

JobClient.runJob(conf);
return 0;
}

public static void main(String[] args) throws Exception {
int res = ToolRunner.run(new Configuration(), new WordCount(), args);
System.exit(res);
}

}

　　2. 保证hadoop集群是配置好了的，单机的也好。新建一个目录，比如 /home/admin/WordCount
　　编译WordCount.java程序。

javac -classpath /home/admin/hadoop/hadoop-0.19.1-core.jar WordCount.java -d /home/admin/WordCount

　　3. 编译完后在/home/admin/WordCount目录会发现三个class文件 WordCount.class，WordCount$Map.class，WordCount$Reduce.class。
　　cd 进入 /home/admin/WordCount目录,然后执行：

jar cvf WordCount.jar *.class

　　就会生成 WordCount.jar 文件。

　　4. 构造一些输入数据
　　input1.txt和input2.txt的文件里面是一些单词。如下：

[admin@host WordCount]$ cat input1.txt
Hello, i love china
are you ok?
[admin@host WordCount]$ cat input2.txt
hello, i love word
You are ok

　　在hadoop上新建目录，和put程序运行所需要的输入文件:

hadoop fs -mkdir /tmp/input
hadoop fs -mkdir /tmp/output
hadoop fs -put input1.txt /tmp/input/
hadoop fs -put input2.txt /tmp/input/

　　5. 运行程序,会显示job运行时的一些信息。

[admin@host WordCount]$ hadoop jar WordCount.jar WordCount /tmp/input /tmp/output
10/09/16 22:49:43 WARN mapred.JobClient: Use GenericOptionsParser for parsing the arguments. Applications should implement Tool for the same.
10/09/16 22:49:43 INFO mapred.FileInputFormat: Total input paths to process :2
10/09/16 22:49:43 INFO mapred.JobClient: Running job: job_201008171228_76165
10/09/16 22:49:44 INFO mapred.JobClient: map 0% reduce 0%
10/09/16 22:49:47 INFO mapred.JobClient: map 100% reduce 0%
10/09/16 22:49:54 INFO mapred.JobClient: map 100% reduce 100%
10/09/16 22:49:55 INFO mapred.JobClient: Job complete: job_201008171228_76165
10/09/16 22:49:55 INFO mapred.JobClient: Counters: 16
10/09/16 22:49:55 INFO mapred.JobClient: File Systems
10/09/16 22:49:55 INFO mapred.JobClient: HDFS bytes read=62
10/09/16 22:49:55 INFO mapred.JobClient: HDFS bytes written=73
10/09/16 22:49:55 INFO mapred.JobClient: Local bytes read=152
10/09/16 22:49:55 INFO mapred.JobClient: Local bytes written=366
10/09/16 22:49:55 INFO mapred.JobClient: Job Counters
10/09/16 22:49:55 INFO mapred.JobClient: Launched reduce tasks=1
10/09/16 22:49:55 INFO mapred.JobClient: Rack-local map tasks=2
10/09/16 22:49:55 INFO mapred.JobClient: Launched map tasks=2
10/09/16 22:49:55 INFO mapred.JobClient: Map-Reduce Framework
10/09/16 22:49:55 INFO mapred.JobClient: Reduce input groups=11
10/09/16 22:49:55 INFO mapred.JobClient: Combine output records=14
10/09/16 22:49:55 INFO mapred.JobClient: Map input records=4
10/09/16 22:49:55 INFO mapred.JobClient: Reduce output records=11
10/09/16 22:49:55 INFO mapred.JobClient: Map output bytes=118
10/09/16 22:49:55 INFO mapred.JobClient: Map input bytes=62
10/09/16 22:49:55 INFO mapred.JobClient: Combine input records=14
10/09/16 22:49:55 INFO mapred.JobClient: Map output records=14
10/09/16 22:49:55 INFO mapred.JobClient: Reduce input records=14

　　6. 查看运行结果

[admin@host WordCount]$ hadoop fs -ls /tmp/output/
Found 2 items
drwxr-x--- - admin admin 0 2010-09-16 22:43 /tmp/output/_logs
-rw-r----- 1 admin admin 102 2010-09-16 22:44 /tmp/output/part-00000
[admin@host WordCount]$ hadoop fs -cat /tmp/output/part-00000
Hello, 1
You 1
are 2
china 1
hello, 1
i 2
love 2
ok 1
ok? 1
word 1
you 1