ChenFolin created MAPREDUCE-6710:
------------------------------------

             Summary: Job CommitterEventHandler waitForValidCommitWindow may 
wait forever
                 Key: MAPREDUCE-6710
                 URL: https://issues.apache.org/jira/browse/MAPREDUCE-6710
             Project: Hadoop Map/Reduce
          Issue Type: Bug
          Components: applicationmaster
    Affects Versions: 2.6.1, 2.5.0, 2.8.0, 3.0.0-alpha1
            Reporter: ChenFolin


The callBack Runnable may run after runOnNextHeartbeat but before wait(), if 
that happens, it will wait forever.

So I think the wait() must set timeout.

CommitterEventHandler#waitForValidCommitWindow
while (now - lastHeartbeatTime > commitWindowMs) {
        rmHeartbeatHandler.runOnNextHeartbeat(new Runnable() {
          @Override
          public void run() {
            synchronized (EventProcessor.this) {
              EventProcessor.this.notify();
            }
          }
        });

        wait();
        lastHeartbeatTime = rmHeartbeatHandler.getLastHeartbeatTime();
        now = context.getClock().getTime();
      }



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

---------------------------------------------------------------------
To unsubscribe, e-mail: mapreduce-dev-unsubscr...@hadoop.apache.org
For additional commands, e-mail: mapreduce-dev-h...@hadoop.apache.org

Reply via email to