远程调用执行Hadoop Map/Reduce

水莹儿

　　在Web项目中，由用户下发任务后，后台服务器远程调用JobTracker所在服务器，运行Map/Reduce更符合B/S架构的习惯。
　　由于网上没有相关资料，所以自己实现了一个，现在分享一下。
　　注：基于Hadoop1.1.2版本
　　转发请注明地址：http://sgq0085.iyunv.com/admin/blogs/1879442
　　一个常见的WordCount如下：

package com.gqshao.hadoop.remote;
import java.io.IOException;
import java.util.*;
import org.apache.hadoop.conf.*;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.io.*;
import org.apache.hadoop.mapreduce.*;
import org.apache.hadoop.mapreduce.lib.input.*;
import org.apache.hadoop.mapreduce.lib.output.*;
import org.apache.hadoop.util.*;
public class WordCount extends Configured implements Tool {
public static class Map extends 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, Context context) throws IOException, InterruptedException {
String line = value.toString();
StringTokenizer tokenizer = new StringTokenizer(line);
while (tokenizer.hasMoreTokens()) {
word.set(tokenizer.nextToken());
context.write(word, one);
}
}
}
public static class Reduce extends Reducer<Text, IntWritable, Text, IntWritable> {
public void reduce(Text key, Iterable<IntWritable> values, Context context) throws IOException, InterruptedException {
int sum = 0;
for (IntWritable val : values) {
sum += val.get();
}
context.write(key, new IntWritable(sum));
}
}
public int run(String[] args) throws Exception {
this.getClass().getResource("/hadoop/");
Configuration conf = getConf();
Job job = new Job(conf);
conf.set("mapred.job.tracker", "192.168.0.128:9001");
conf.set("fs.default.name", "hdfs://192.168.0.128:9000");
conf.set("hadoop.job.ugi", "hadoop");
conf.set("Hadoop.tmp.dir", "/user/gqshao/temp/");
job.setJarByClass(WordCount.class);
job.setJobName("wordcount");
job.setOutputKeyClass(Text.class);
job.setOutputValueClass(IntWritable.class);
job.setMapperClass(Map.class);
job.setReducerClass(Reduce.class);
job.setInputFormatClass(TextInputFormat.class);
job.setOutputFormatClass(TextOutputFormat.class);
String hdfs = "hdfs://192.168.0.128:9000";
args = new String[] { hdfs + "/user/gqshao/input/big", hdfs + "/user/gqshao/output/WordCount/" + new Date().getTime() };
FileInputFormat.setInputPaths(job, new Path(args[0]));
FileOutputFormat.setOutputPath(job, new Path(args[1]));
boolean success = job.waitForCompletion(true);
return success ? 0 : 1;
}
public static void main(String[] args) throws Exception {
int ret = ToolRunner.run(new WordCount(), args);
System.exit(ret);
}
}

 在这里输入和输出目录都是指向HDFS上的，但实际运行的时候(一般 -Xms128m -Xmx512m -XX:MaxPermSize=128M)发现输出中有如下信息：
信息: Running job: job_local_0001
　　证明该Map/Reduce程序运行在Local中。也就是说，这种方式只能提前打好Jar包，放到Cluster服务器上，在通过Jar运行。
　　转发请注明地址：http://sgq0085.iyunv.com/admin/blogs/1879442
　　如何远程运行Map/Reduce程序，经研究发现两点。
　　1.需要将Hadoop的配置文件加载到当前进程的ClassLoader中，或将配置文件放到/bin目录下。
　　通过跟踪 job.waitForCompletion(true);→submit();→info = jobClient.submitJobInternal(conf);→status = jobSubmitClient.submitJob(jobId, submitJobDir.toString(), jobCopy.getCredentials());
　　发现private JobSubmissionProtocol jobSubmitClient;分别有两个实现
　　在org.apache.hadoop.mapred.JobClient中init()方法中可以看到如果设置了conf中如果设置了mapred.job.tracker则在Hadoop Cluster中运行，否则是Local

  public void init(JobConf conf) throws IOException {
String tracker = conf.get("mapred.job.tracker", "local");
tasklogtimeout = conf.getInt(
TASKLOG_PULL_TIMEOUT_KEY, DEFAULT_TASKLOG_TIMEOUT);
this.ugi = UserGroupInformation.getCurrentUser();
if ("local".equals(tracker)) {
conf.setNumMapTasks(1);
this.jobSubmitClient = new LocalJobRunner(conf);
} else {
this.rpcJobSubmitClient =
createRPCProxy(JobTracker.getAddress(conf), conf);
this.jobSubmitClient = createProxy(this.rpcJobSubmitClient, conf);
}        
}
　　所以需要在运行时加载某目录下配置文件
　　方法如下：

    /**
* 加载配置文件
*/
public static void setConf(Class<?> clazz, Thread thread, String path) {
URL url = clazz.getResource(path);
try {
File confDir = new File(url.toURI());
if (!confDir.exists()) {
return;
}
URL key = confDir.getCanonicalFile().toURI().toURL();
ClassLoader classLoader = thread.getContextClassLoader();
classLoader = new URLClassLoader(new URL[] { key }, classLoader);
thread.setContextClassLoader(classLoader);
} catch (Exception e) {
e.printStackTrace();
}
}
　　2.设置运行时Jar包
　　继续看jobClient.submitJobInternal(conf);可以发现client在提交作业到Hadoop时需要把作业打包成jar，然后copy到fs的submitJarFile路径中。所以必须指定conf中的运行的Jar包。
　　方法如下：

    /**
* 动态生成Jar包
*/
public static File createJar(Class<?> clazz) throws Exception {
String fqn = clazz.getName();
String base = fqn.substring(0, fqn.lastIndexOf("."));
base = "/" + base.replaceAll("\\.", Matcher.quoteReplacement("/"));
URL root = clazz.getResource("");
JarOutputStream out = null;
final File jar = File.createTempFile("HadoopRunningJar-", ".jar", new File(System.getProperty("java.io.tmpdir")));
System.out.println(jar.getAbsolutePath());
Runtime.getRuntime().addShutdownHook(new Thread() {
public void run() {
jar.delete();
}
});
try {
File path = new File(root.toURI());
Manifest manifest = new Manifest();
manifest.getMainAttributes().putValue("Manifest-Version", "1.0");
manifest.getMainAttributes().putValue("Created-By", "RemoteHadoopUtil");
out = new JarOutputStream(new FileOutputStream(jar), manifest);
writeBaseFile(out, path, base);
} finally {
out.flush();
out.close();
}
return jar;
}
/**
* 递归添加.class文件
*/
private static void writeBaseFile(JarOutputStream out, File file, String base) throws IOException {
if (file.isDirectory()) {
File[] fl = file.listFiles();
if (base.length() > 0) {
base = base + "/";
}
for (int i = 0; i < fl.length; i++) {
writeBaseFile(out, fl, base + fl.getName());
}
} else {
out.putNextEntry(new JarEntry(base));
FileInputStream in = null;
try {
in = new FileInputStream(file);
byte[] buffer = new byte[1024];
int n = in.read(buffer);
while (n != -1) {
out.write(buffer, 0, n);
n = in.read(buffer);
}
} finally {
in.close();
}  
}
}
　　修改后的WordCount如下：

public class WordCount extends Configured implements Tool {
public static class Map extends 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, Context context) throws IOException, InterruptedException {
String line = value.toString();
System.out.println("line===>" + line);
StringTokenizer tokenizer = new StringTokenizer(line);
while (tokenizer.hasMoreTokens()) {
word.set(tokenizer.nextToken());
context.write(word, one);
}
}
}
public static class Reduce extends Reducer<Text, IntWritable, Text, IntWritable> {
public void reduce(Text key, Iterable<IntWritable> values, Context context) throws IOException, InterruptedException {
int sum = 0;
for (IntWritable val : values) {
sum += val.get();
}
context.write(key, new IntWritable(sum));
}
}
public int run(String[] args) throws Exception {
Configuration conf = getConf();
Job job = new Job(conf);
System.out.println(conf.get("mapred.job.tracker"));
System.out.println(conf.get("fs.default.name"));
/**
* TODO:调用二
*/
File jarFile = RemoteHadoopUtil.createJar(WordCount.class);
((JobConf) job.getConfiguration()).setJar(jarFile.toString());
job.setJarByClass(WordCount.class);
job.setJobName("wordcount");
job.setOutputKeyClass(Text.class);
job.setOutputValueClass(IntWritable.class);
job.setMapperClass(Map.class);
job.setReducerClass(Reduce.class);
job.setInputFormatClass(TextInputFormat.class);
job.setOutputFormatClass(TextOutputFormat.class);
String hdfs = "hdfs://192.168.0.128:9000";
args = new String[] { hdfs + "/user/gqshao/input/WordCount/", hdfs + "/user/gqshao/output/WordCount/" + new Date().getTime() };
FileInputFormat.setInputPaths(job, new Path(args[0]));
FileOutputFormat.setOutputPath(job, new Path(args[1]));
boolean success = job.waitForCompletion(true);
System.out.println(job.isComplete());
System.out.println("JobID: " + job.getJobID());
return success ? 0 : 1;
}
public static void main(String[] args) throws Exception {
/**
* TODO:调用一
*/
RemoteHadoopUtil.setConf(WordCount.class, Thread.currentThread(), "/hadoop");
int ret = ToolRunner.run(new WordCount(), args);
System.exit(ret);
}
}
　　转发请注明地址：http://sgq0085.iyunv.com/admin/blogs/1879442
 附件中有完整代码和测试用例，欢迎讨论。解压后在文件目录中运行mvn eclipse:clean eclipse:eclipse即可(前提是需要有Maven)