hadoop中mapreduce如何实现串联执行

小编给大家分享一下hadoop中mapreduce如何实现串联执行，相信大部分人都还不怎么了解，因此分享这篇文章给大家参考一下，希望大家阅读完这篇文章后大有收获，下面让我们一起去了解一下吧！

import java.io.IOException;
import java.util.Iterator;
import org.apache.commons.logging.Log;
import org.apache.commons.logging.LogFactory;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.io.LongWritable;
import org.apache.hadoop.io.NullWritable;
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.jobcontrol.ControlledJob;
import org.apache.hadoop.mapreduce.lib.jobcontrol.JobControl;
import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;
public class PickMain {
	private static final Log LOG = LogFactory.getLog(PickMain.class);
	public static void main(String[] args) throws IOException, ClassNotFoundException, InterruptedException {
/*		
 * Configuration conf = new Configuration();
		Job job1 = Job.getInstance(conf);
		
		job1.setJarByClass(PickMain.class);
		job1.setMapperClass(FindMapper.class);
		job1.setReducerClass(FindReducer.class);
		job1.setOutputKeyClass(Text.class);
		job1.setOutputValueClass(Text.class);
		
		FileInputFormat.addInputPath(job1, new Path(args[0]));
		FileOutputFormat.setOutputPath(job1, new Path(args[1]));
		
		boolean flag1 = job1.waitForCompletion(true);
		//下面这种方法也可以实现串联执行job
		if(flag1) {
			Job job2 = Job.getInstance(conf);
	
			job2.setJarByClass(PickMain.class);
			job2.setMapperClass(SecondFindMapper.class);
			job2.setReducerClass(SecondFindReducer.class);
			job2.setOutputKeyClass(Text.class);
			job2.setOutputValueClass(Text.class);
			
			FileInputFormat.addInputPath(job2, new Path(args[1]));
			FileOutputFormat.setOutputPath(job2, new Path(args[2]));
			
			boolean flag2 = job2.waitForCompletion(true);
			System.out.println(flag2?0:1);
			if(flag2) {
				LOG.info("The job is done!");				
				System.exit(0);
			}else {
				LOG.info("The Second job is wrong!");
				System.exit(1);
			}
						
	 }else {
			LOG.info("The firt job is Running Wrong  job break!");
			System.exit(1);
		}
		
		*/
		
		
		//下面通过使用ContolledJob和JobControl来实现提交多个作业
		
		
		Configuration conf = new Configuration();
		Job job1 = Job.getInstance(conf);
		
		job1.setJarByClass(PickMain.class);
		job1.setMapperClass(FindMapper.class);
		job1.setReducerClass(FindReducer.class);
		job1.setOutputKeyClass(Text.class);
		job1.setOutputValueClass(Text.class);
		
		FileInputFormat.addInputPath(job1, new Path(args[0]));
		FileOutputFormat.setOutputPath(job1, new Path(args[1]));
		
		Configuration conf2 = new Configuration();
		
		Job job2 = Job.getInstance(conf2);
		
		job2.setJarByClass(PickMain.class);
		job2.setMapperClass(SecondFindMapper.class);
		job2.setReducerClass(SecondFindReducer.class);
		job2.setOutputKeyClass(Text.class);
		job2.setOutputValueClass(Text.class);
		
		FileInputFormat.addInputPath(job2, new Path(args[1]));
		FileOutputFormat.setOutputPath(job2, new Path(args[2]));
		//创建ControlledJob对job进行包装
		ControlledJob cjob1 = new ControlledJob(conf);
		ControlledJob cjob2 = new ControlledJob(conf2);
		cjob1.setJob(job1);
		cjob2.setJob(job2);
		//设置依赖关系，这个时候只有等到job1执行完成后job2才会执行
		cjob2.addDependingJob(cjob1);
		
		//JobControl该类相当于一个job控制器，它是一个线程，需要通过线程启动
		JobControl jc = new JobControl("my_jobcontrol");
		jc.addJob(cjob1);
		jc.addJob(cjob2);
		Thread th = new Thread(jc);
		th.start();
		//等到所有的job都执行完成后在退出
		while(!jc.allFinished()) {
			Thread.sleep(5000);
		}
		System.exit(0);
		
	}
}
class FindMapper extends Mapper<LongWritable, Text, Text, Text>{
	Text m1 = new Text();
	Text m2 = new Text();
	@Override
	protected void map(LongWritable key, Text value, Mapper<LongWritable, Text, Text, Text>.Context context)
			throws IOException, InterruptedException {
		String line = value.toString();
		String[] tmp1 = line.split(":");
		String outval = tmp1[0];
		String[] outkeys = tmp1[1].split(",");
		for(int i = 0 ; i<outkeys.length;i++) {
			m1.set(outkeys[i]);m2.set(outval);
			context.write(m1,m2);
		}	
	}	
}
class FindReducer extends Reducer<Text, Text, Text, NullWritable>{
	StringBuilder sb = new StringBuilder();
	NullWritable nul = NullWritable.get();
	Text outval = new Text();
	String spector = ":";
	@Override
	protected void reduce(Text txt, Iterable<Text> txtiter, Reducer<Text, Text, Text, NullWritable>.Context context)
			throws IOException, InterruptedException {
		sb.delete(0, sb.length());
		sb.append(txt.toString());
		Iterator<Text> it = txtiter.iterator();
		while(it.hasNext()) {
			sb.append(spector+it.next().toString());
		}
		outval.set(sb.toString());
		context.write(outval, nul);
	}
	
}
class SecondFindMapper extends Mapper<LongWritable, Text, Text, Text>{
	Text keyout = new Text();
	Text valueout = new Text();
	@Override
	protected void map(LongWritable key, Text value, Mapper<LongWritable, Text, Text, Text>.Context context)
			throws IOException, InterruptedException {
		String[] fs = value.toString().split(":");
		valueout.set(fs[0]);
		if(fs.length>0) {
			for(int i = 1;i<fs.length-1;i++) {
				for(int j = i+1;j<fs.length;j++) {
					if((int)fs[i].toCharArray()[0]>(int)fs[j].toCharArray()[0]) {
						keyout.set(fs[j]+"-"+fs[i]);
					}else {
						keyout.set(fs[i]+"-"+fs[j]);
					}						
					context.write(keyout, valueout);
					
				}
			}
			
		}		
	}	
}
class  SecondFindReducer extends Reducer<Text, Text, Text, Text>{
	StringBuilder sb = new StringBuilder();
	Text outvalue = new Text();
	@Override
	protected void reduce(Text key, Iterable<Text> iter, Reducer<Text, Text, Text, Text>.Context context)
			throws IOException, InterruptedException {
		sb.delete(0, sb.length());
		Iterator<Text> it =  iter.iterator();
		if(it.hasNext()) {
			sb.append(it.next().toString());
		}
		
		while(it.hasNext()) {
			sb.append(","+it.next().toString());
		}
		outvalue.set(sb.toString());
		context.write(key, outvalue);		
	}	
}

以上是“hadoop中mapreduce如何实现串联执行”这篇文章的所有内容，感谢各位的阅读！相信大家都有了一定的了解，希望分享的内容对大家有所帮助，如果还想学习更多知识，欢迎关注亿速云行业资讯频道！