MapReduce编程场景之多Job串联

MapReduce编程场景之多Job串联

（一）需求
一个稍复杂点的处理逻辑往往需要多个 MapReduce 程序串联处理，多 job 的串联可以借助MapReduce 框架的 JobControl 实现

（二）分析
有两个 MapReduce 任务，分别是 Flow 的 SumMR 和 SortMR，其中有依赖关系：SumMR的输出是 SortMR 的输入，所以 SortMR 的启动得在 SumMR 完成之后
（三）代码实现

/**
*@author lv_hulk
*/
public static void main(String[] args) throws Exception {
Configuration conf = new Configuration();
Job jobsum = Job.getInstance(conf);
jobsum.setJarByClass(RunManyJobMR.class);
jobsum.setMapperClass(FlowSumMapper.class);
Stay hungry Stay foolish -- http://blog.csdn.net/zhongqi2513
jobsum.setReducerClass(FlowSumReducer.class);
jobsum.setMapOutputKeyClass(Text.class);
jobsum.setMapOutputValueClass(Flow.class);
jobsum.setCombinerClass(FlowSumReducer.class);
jobsum.setOutputKeyClass(Text.class);
jobsum.setOutputValueClass(Text.class);
FileInputFormat.setInputPaths(jobsum, "d:/flow/input");
FileOutputFormat.setOutputPath(jobsum, new Path("d:/flow/output12"));
Job jobsort = Job.getInstance(conf);
jobsort.setJarByClass(RunManyJobMR.class);
jobsort.setMapperClass(FlowSortMapper.class);
jobsort.setReducerClass(FlowSortReducer.class);
jobsort.setMapOutputKeyClass(Flow.class);
jobsort.setMapOutputValueClass(Text.class);
jobsort.setOutputKeyClass(NullWritable.class);
jobsort.setOutputValueClass(Flow.class);
FileInputFormat.setInputPaths(jobsort, "d:/flow/output12");
FileOutputFormat.setOutputPath(jobsort, new Path("d:/flow/sortoutput12"));
ControlledJob sumcj = new ControlledJob(jobsum.getConfiguration());
ControlledJob sortcj = new ControlledJob(jobsort.getConfiguration());
sumcj.setJob(jobsum);
sortcj.setJob(jobsort);
// 设置作业依赖关系
sortcj.addDependingJob(sumcj);
JobControl jc = new JobControl("flow sum and sort");
jc.addJob(sumcj);
jc.addJob(sortcj);
Thread jobThread = new Thread(jc);
jobThread.start();
while(!jc.allFinished()){
Thread.sleep(500);
}
jc.stop();
}