[jira] [Commented] (YARN-275) Make NodeManagers to NOT blindly heartbeat irrespective of whether previous heartbeat is processed or not.

["Siddharth Seth (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/YARN-275?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13569115#comment-13569115 ] 

Siddharth Seth commented on YARN-275:
-------------------------------------

bq. To be clear, I did not mean we should implement dynamic heartbeat intervals. Just change the protocol so that heartbeat comes from the RM to the NM. The dynamic part can be done later. I think the patch in this jira (after some minor polishing) limits itself to just the protocol changes as a first step. The RM sends a fixed value obtained via config.
Misunderstood your previous comment to include a dynamic back off. Having the RM dictate the heartbeat interval seems reasonably independent of changes in YARN-365. As an initial step, do you think this will be simpler via config - instead of having the scheduler indicate whether the NMs should back off (like is done in the current patch)? 
                
> Make NodeManagers to NOT blindly heartbeat irrespective of whether previous heartbeat is processed or not.
> ----------------------------------------------------------------------------------------------------------
>
>                 Key: YARN-275
>                 URL: https://issues.apache.org/jira/browse/YARN-275
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: nodemanager, resourcemanager
>            Reporter: Vinod Kumar Vavilapalli
>            Assignee: Xuan Gong
>         Attachments: Prototype.txt, YARN-270.1.patch
>
>
> We need NMs to back off. The event handler mechanism is very scalable but not infinitely so :)

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira