[
https://issues.apache.org/jira/browse/HADOOP-4880?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12658008#action_12658008
]
Devaraj Das commented on HADOOP-4880:
-------------------------------------
The patch still assumes that job completions are ordered. I think we should
remove this assumption. Also, if this particular testcase
testRecoveryWithMultipleJobs may actually be removed. The only additional thing
it checks for is the priority change of a job after a restart. That might be
moved to some other testcase.
> Improvements to TestJobTrackerRestart
> -------------------------------------
>
> Key: HADOOP-4880
> URL: https://issues.apache.org/jira/browse/HADOOP-4880
> Project: Hadoop Core
> Issue Type: Improvement
> Components: mapred, test
> Reporter: Arun C Murthy
> Assignee: Amar Kamat
> Attachments: HADOOP-4880-v1.patch
>
>
> TestJobTrackerRestart could use the following improvements:
> # Remove the one minute 'wait' - this is really bad for test-cases
> # It assumes that a particular ordering of job-scheduling based on current
> behaviour of JobQueueTaskScheduler assigning a single task per heartbeat (pre
> HADOOP-3136) and thus a particular job completion order, which is incorrect
> and won't work with all schedulers.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
