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 Tue, 07 Jan 2014 02:48:51 GMT

    [ https://issues.apache.org/jira/browse/YARN-1506?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13863854#comment-13863854

Junping Du commented on YARN-1506:

Thanks Jian for review and great comments! Please see my reply:
bq. What about other node states other than RUNNING, is RESOURCE_UPDATE event never possible
to come at those states?
Very good point. I think I missed other cases before. Basically, now I want to allow node
in RUNNING, NEW and REBOOT state to be updated with resource. And node in unusable state {UNHEALTHY,
DECOMMISSIONED, LOST} will log/throw exception. Make sense? 
Also, I am curious on it looks like we are missing some transitions, i.e.: 
REBOOT -> RUNNING for a rebooted node come back as running for accepting RECONNECTED/CLEAN_CONTAINER/APP
DECOMMISSIONED -> RUNNING for a decommissioned node be recommissoned again 
LOST -> NEW/UNHELATHY/DECOMMISSONED for a expired node heartbeat again
UNHEALTHY -> RUNNING for a unhealthy node report to be healthy again 
Am I missing anything here?
bq. ResourceOption.build() is not used anywhere?
Nice catch! Will remove it in next patch.
bq. Maybe send the NODE_RESOURCE_UPDATE event from RMNode instead from AdminService? as inside
SchedulerEventType, it’s actually saying the source is from node.
I thought this way before as following common practice. However, given we agreed that some
information (OvercommitTimeout) is unnecessary to go to RMNode and cached there. I think it
could be better to send separated events to RMNode and Scheduler by AdminService. Thoughts?

> 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
>            Priority: Blocker
>         Attachments: YARN-1506-v1.patch, YARN-1506-v2.patch, YARN-1506-v3.patch, YARN-1506-v4.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

View raw message