hadoop jobId

JobClient提交作业前向JobTracker申请一个新的JobId,   job_201107131247_575464
JobTracker:
  // Allocates a new JobId string.
  public synchronized JobID getNewJobId() throws IOException {
    return new JobID(getTrackerIdentifier(), nextJobId++);
  }

private int nextJobId = 1; // nextJobId 是JobTracker的一个内部变量，初始时为1，每new一个JobID加1.

// Get the unique identifier (ie. timestamp) of this job tracker start.
public String getTrackerIdentifier() { //jobtracker启动时创建的一个唯一标识，用启动时的timestamp.
    return trackerIdentifier;
  }

private String trackerIdentifier;  // trackerIdentifier 是JobTracker的一个内部变量，赋值是在构造函数中：
JobTracker(JobConf conf) ｛
    trackerIdentifier = getDateFormat().format(new Date());
｝

  private static SimpleDateFormat getDateFormat() {
    return new SimpleDateFormat("yyyyMMddHHmm");
  }

org.apache.hadoop.mapred.JobID
/**
* JobID represents the immutable and unique identifier for
* the job. JobID consists of two parts. First part
* represents the jobtracker identifier, so that jobID to jobtracker map
* is defined. For cluster setup this string is the jobtracker
* start time, for local setting, it is "local".
* Second part of the JobID is the job number. <br>
* An example JobID is :
* <code>job_200707121733_0003</code> , which represents the third job
* running at the jobtracker started at <code>200707121733</code>.
* <p>
* Applications should never construct or parse JobID strings, but rather
* use appropriate constructors or {@link #forName(String)} method.
*
* @see TaskID
* @see TaskAttemptID
*/


hadoop ： 一个文件对应有一个.crc的校验和文件，如果这个校验和文件有问题，那么hadoop fs -put的时候就会报错。方法是删除这个校验和文件。
-rwxrwxrwx   1 c users       5 07-13 17:06 data.dat
-rw-rw-r--   1 c users    8536 08-22 20:25 .data.dat.crc