[ http://issues.apache.org/jira/browse/HADOOP-813?page=comments#action_12457857 ] Owen O'Malley commented on HADOOP-813: --------------------------------------
1. sortComplete should be declared volatile. 2. The sortProgress thread should use LOG.info for reporting exceptions rather than System.out.println. 3. I'd probably add a sortProgress.interrupt() when you are closing down so that it can shut down faster. 4. The sortProgress thread should be a daemon thread. 5. There seem to be a lot of other changes in this patch that aren't related to the issue and aren't mentioned in your comment. Are they intentional? > map tasks lost during sort > -------------------------- > > Key: HADOOP-813 > URL: http://issues.apache.org/jira/browse/HADOOP-813 > Project: Hadoop > Issue Type: Bug > Components: mapred > Affects Versions: 0.10.0 > Reporter: Owen O'Malley > Assigned To: Devaraj Das > Fix For: 0.10.0 > > Attachments: 813.patch > > > On a 500 node cluster, I had a bunch of map tasks get "lost" because they > failed to report progress for 10 minutes. They appear to be in the sort stage > at the end of the map. I hypothesize that the patch for HADOOP-331 does not > update the map's progress during the sort/merge. If the sort/merge takes more > than 10 minutes, the task is lost. -- This message is automatically generated by JIRA. - If you think it was sent incorrectly contact one of the administrators: http://issues.apache.org/jira/secure/Administrators.jspa - For more information on JIRA, see: http://www.atlassian.com/software/jira
