You are viewing a plain text version of this content. The canonical link for it is here.
Posted to mapreduce-dev@hadoop.apache.org by "Amar Kamat (JIRA)" <ji...@apache.org> on 2009/07/08 07:30:16 UTC

[jira] Created: (MAPREDUCE-731) Factor out the timer code in jobtracker and add a callback interface

Factor out the timer code in jobtracker and add a callback interface
--------------------------------------------------------------------

                 Key: MAPREDUCE-731
                 URL: https://issues.apache.org/jira/browse/MAPREDUCE-731
             Project: Hadoop Map/Reduce
          Issue Type: Improvement
          Components: jobtracker
            Reporter: Amar Kamat


I can see LostTracker, ExpiryJobs and ExpiryLaunchingTasks code in the jobtracker which essentially do the same thing. I see a better way of doing this by having a time utility which simply waits for the object to expire. Once expired, it invoked the master via the callback interface. So the jobtracker code would like
{code}
void expire(ExpiryItem item) {
  if (item instanceOf JobInProgress) {
    // expire job
  } else if (item instanceOf TaskTracker) {
    // expire tracker ... i.e lost tracker
  } else if (item instanceOf TaskAttempt) {
     // expire attempt
  }

  // create a time thread for each code piece 
  losttracker-timer = new Timer();
  expiry-timer = new Timer();
  attempt-timer = new Timer();

  // any special logic should be implemented before adding to the expiry queue.
}
{code}

JobInProgress, TaskAttempt and TaskTracker all should implement ExpiryItem interface (which is nothing but getTime() or something similar). 

This will help us in 2 ways
1) Maintenance : 
2) Testing :
 
Thoughts?

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.