hadoop-common-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Hemanth Yamijala (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HADOOP-5739) After JobTracker restart Capacity Schduler does not schedules pending tasks from already running tasks.
Date Mon, 25 May 2009 04:56:45 GMT

    [ https://issues.apache.org/jira/browse/HADOOP-5739?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12712636#action_12712636
] 

Hemanth Yamijala commented on HADOOP-5739:
------------------------------------------

Some comments:

- In JobTracker.processTaskAttempt, we are getting TaskAttemptID.forName(taskAttemptId) again.
Can't we use the instance that's already obtained in the beginning of this method.
- MAX_COMPLETE_USER_JOBS_IN_MEMORY is being hardcoded to 0. I think this is a mistake, maybe
for testing ?
- I think it is good to log that a status change event is being raised at an info level.
- shouldn't the call to updateJobInProgressListeners be synchronized on JT instance. It is
everywhere else.
- Since the bug is not in Capacity scheduler, it looks like this is not required to me.
- TestJobInProgressListener: The new test will cover the code path added only if there is
a taskattempt id record for the setup task in the history file. I don't see how we can guarantee
that. Is there any way ?
- I think this should use the same conf parameters and same job parameters as TestJobTrackerRestart.
That could be one way of ensuring that the setup task attempt is written to the history file.
- We should verify the events expected are raised. Not doing that.
- We seem to be starting datanodes twice.
- Comment "// Turn off the recovery" should be "turn on.."
- TmpJSListener -> Rename as FakeJobStatusListener.
- TmpScheduler -> FakeTaskScheduler.



> After JobTracker restart  Capacity Schduler does not schedules pending tasks from already
running tasks.
> --------------------------------------------------------------------------------------------------------
>
>                 Key: HADOOP-5739
>                 URL: https://issues.apache.org/jira/browse/HADOOP-5739
>             Project: Hadoop Core
>          Issue Type: Bug
>          Components: contrib/capacity-sched
>    Affects Versions: 0.20.0
>            Reporter: Karam Singh
>            Assignee: rahul k singh
>         Attachments: hadoop-5739-latest.patch, hadoop-5739.patch, hadoop-5739.patch
>
>
> After JobTracker restart  Capacity Schduler does not schedules pending task from already
running tasks.

-- 
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