hadoop-mapreduce-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Hong Tang (JIRA)" <j...@apache.org>
Subject [jira] Commented: (MAPREDUCE-986) rumen makes a task with a null type when one of the task lines is truncated
Date Tue, 15 Sep 2009 22:37:57 GMT

    [ https://issues.apache.org/jira/browse/MAPREDUCE-986?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12755755#action_12755755
] 

Hong Tang commented on MAPREDUCE-986:
-------------------------------------

{noformat}
-      task.setTaskType(typ);
+      if (!taskAlreadyLogged || typ != null) {
+        task.setTaskType(typ);
+      }
{noformat}
Shouldn't it be just:
{noformat}
-      task.setTaskType(typ);
+      if (typ != null) {
+        task.setTaskType(typ);
+      }
{noformat}
Otherwise, you would still the task type if taskAlreadyLogged==true but typ is null, and then
later be ignored.

Also -1 on modifying existing tests to test bug fixes. It'd be better to have a standalone
negative test case to test this logic instead of piggybacking the testing with another test.

> rumen makes a task with a null type when one of the task lines is truncated
> ---------------------------------------------------------------------------
>
>                 Key: MAPREDUCE-986
>                 URL: https://issues.apache.org/jira/browse/MAPREDUCE-986
>             Project: Hadoop Map/Reduce
>          Issue Type: Bug
>            Reporter: Dick King
>            Assignee: Dick King
>         Attachments: MAPREDUCE-986--2009-09-15--1518.patch
>
>
> Rumen was used to produce a job trace, but the job trace contained a LoggedTask that
had a null taskType.  This appears to happen when a Task line is truncated.
> We should not put the LoggedTask in the trace at all when this happens.

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


Mime
View raw message