hadoop-yarn-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Junping Du (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (YARN-1506) Replace set resource change on RMNode/SchedulerNode directly with event notification.
Date Thu, 14 Aug 2014 01:40:13 GMT

    [ https://issues.apache.org/jira/browse/YARN-1506?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14096407#comment-14096407
] 

Junping Du commented on YARN-1506:
----------------------------------

Thanks [~bikassaha] and [~jianhe] for review and comments!
bq. At this point the previously accepted 4GB request cannot be satisfied and the application
will get stuck. We may need to follow this up in a different jira.
Agree. Agree to discuss this in a separate JIRA. Improve unit tests as your previous comments.


bq. What is more important is that having a shared method declares that these pieces of code
are related to each other in a logical way (if such a relation exists). 
Done. Abstract to a common method.

bq. Since it’s changed to asynchronous, we may change the log to not say successfully.
Removed.

bq. IMO, since UpdateNodeResourceWhenUnusableTransition and UpdateNodeResourceWhenNonRunningTransition
are the same except one extra logging, we can do the logging for both and just keep one transition?
Agree. Merged.

bq. if possible, nodeResourceUpdate method can be moved into AbstractYarnScheduler, a new
common base class for sharing common code among all the schedulers.
That's good point. Move to AbstractYarnScheduler now.

bq. SchedulerNode.setTotalResource -> SchedulerNode.updateTotalAndAvailableResource() ?
My current thinking is better to keep it as setTotalResource, as the later one may mistake
people that we can set availableResource as a separate value in this method. Actually, we
only set totalResource and availableResource is just get refresh to keep consistent with totalResource
change. May be we can keep it here?

bq. UpdateNodeResourceResponse should be an abstract class which implements newInstance()
method.
Fixed.

bq. AdminService.updateNodeResource should RMAuditLogger to log the operations as well.
Fixed.

> Replace set resource change on RMNode/SchedulerNode directly with event notification.
> -------------------------------------------------------------------------------------
>
>                 Key: YARN-1506
>                 URL: https://issues.apache.org/jira/browse/YARN-1506
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: nodemanager, scheduler
>            Reporter: Junping Du
>            Assignee: Junping Du
>         Attachments: YARN-1506-v1.patch, YARN-1506-v10.patch, YARN-1506-v11.patch, YARN-1506-v2.patch,
YARN-1506-v3.patch, YARN-1506-v4.patch, YARN-1506-v5.patch, YARN-1506-v6.patch, YARN-1506-v7.patch,
YARN-1506-v8.patch, YARN-1506-v9.patch
>
>
> According to Vinod's comments on YARN-312 (https://issues.apache.org/jira/browse/YARN-312?focusedCommentId=13846087&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-13846087),
we should replace RMNode.setResourceOption() with some resource change event.



--
This message was sent by Atlassian JIRA
(v6.2#6252)

Mime
View raw message