客户端MapReduce提交到YARN过程

在Mapreduce v1中是使用JobClient来和JobTracker交互完成Job的提交，用户先创建一个Job，通过JobConf设置好参数，通过JobClient提交并监控Job的进展，在JobClient中有一个内部成员变量JobSubmissionProtocol，JobTracker实现了该接口，通过该协议客户端和JobTracker通信完成作业的提交

  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();
    //如果mapred.job.tracker设置成local，则创建本地LocalJobRunner，否则创建RPC代理
    if ("local".equals(tracker)) {
      conf.setNumMapTasks(1);
      this.jobSubmitClient = new LocalJobRunner(conf);
    } else {
      this.jobSubmitClient = createRPCProxy(JobTracker.getAddress(conf), conf);
    }        
  }

按顺序调用:

Job.waitForCompletion()

Job.submit()

jobClient.submitJobInternal()

jobSubmitClient.submitJob(jobId, submitJobDir.toString(), jobCopy.getCredentials())

完成作业提交

而YARN的作业提交procotol是ClientRMProtocol，提交MRv2作业时，首先会生成集群信息类cluster，里面有一个frameworkLoader内部变量会从配置文件中加载ClientProtocolProvider的实现类，这里分别是LocalClientProtocolProvider和 YarnClientProtocolProvider 。Cluster类在initialize中，会遍历frameworkLoader，由ClientProtocolProvider来生成具体的ClientProtocol ，比如在YarnClientProtocolProvider中就会判断JobConf中的 mapreduce.framework.name是否为 yarn，如果是的话则会生成YARNRunner

YarnClientProtocolProvider的create方法:

  @Override
  public ClientProtocol create(Configuration conf) throws IOException {
    if (MRConfig.YARN_FRAMEWORK_NAME.equals(conf.get(MRConfig.FRAMEWORK_NAME))) {
      return new YARNRunner(conf);
    }
    return null;
  }

ClientProtocol目前有两个实现 YARNRunner 和LocalJobRunner，LocalJobRunner(mapreduce.framework.name为local )主要是在本地执行mapreduce，可以方便对程序进行调试。YARNRunner是将作业提交到YARN上。

YARNRunner初始化会和ResourceManager建立RPC链接（默认是8032端口），真正和RM通信的协议是 ClientRMProtocol ，客户端和RM交互的所有操作都会通过YARNRunner的成员变量 rmClient( ClientRMProtocol )提交出去，比如killApplication, getNodeReports, getJobCounters等等

  public synchronized void start() {
    YarnRPC rpc = YarnRPC.create(getConfig());
    this.rmClient = (ClientRMProtocol) rpc.getProxy(
        ClientRMProtocol.class, rmAddress, getConfig());
    if (LOG.isDebugEnabled()) {
      LOG.debug("Connecting to ResourceManager at " + rmAddress);
    }
    super.start();
  }

Cluster类初始化完成后，就要生成Application了，先和RM通信申请一个Application(getNewApplication )，得到一个GetNewApplicationResponse，里面封装了ApplicationID，和RM能提供的最小、最大Resource Capacity

public interface GetNewApplicationResponse {
  public abstract ApplicationId getApplicationId();
  public Resource getMinimumResourceCapability();
  public Resource getMaximumResourceCapability();
  public void setMaximumResourceCapability(Resource capability); 
}

Resource定义了一组集群计算资源，目前只把memory和cpu纳入进来，这边的cpu指virtual core，也就是一个物理core可以被认为抽象成多个virtual core，而非一对一对应关系

public abstract class Resource implements Comparable<Resource> {
  public abstract int getMemory();
  public abstract void setMemory(int memory);
  public abstract int getVirtualCores();
  public abstract void setVirtualCores(int vCores);
}

然后需要构造ApplicationSubmissionContext，其中包含了启动MR AM的信息，比如提交的job在HDFS的staging目录路径（job.xml, job.split, job.splitmetainfo, libjars, files, archives等），用户ugi信息，Secure Tokens。完成context构造后，调用resMgrDelegate.submitApplication(appContext)

YARNRunner的submitJob方法:

  @Override
  public JobStatus submitJob(JobID jobId, String jobSubmitDir, Credentials ts)
  throws IOException, InterruptedException {
    // Construct necessary information to start the MR AM
    ApplicationSubmissionContext appContext =
      createApplicationSubmissionContext(conf, jobSubmitDir, ts);

    // Submit to ResourceManager
    ApplicationId applicationId = resMgrDelegate.submitApplication(appContext);

    ApplicationReport appMaster = resMgrDelegate.getApplicationReport(applicationId);
    String diagnostics = (appMaster == null ?
            "application report is null" : appMaster.getDiagnostics());
    if (appMaster == null || appMaster.getYarnApplicationState() == YarnApplicationState.FAILED
        || appMaster.getYarnApplicationState() == YarnApplicationState.KILLED) {
      throw new IOException("Failed to run job : " +
        diagnostics);
    }
    return clientCache.getClient(jobId).getJobStatus(jobId);
  }

最后通过getJobStatus方法获得Job状态信息

    org.apache.hadoop.mapreduce.v2.api.records.JobId jobId =
      TypeConverter.toYarn(oldJobID);
    GetJobReportRequest request =
        recordFactory.newRecordInstance(GetJobReportRequest.class);
    request.setJobId(jobId);
    JobReport report = ((GetJobReportResponse) invoke("getJobReport",
        GetJobReportRequest.class, request)).getJobReport();