Zhijie Shen commented on YARN-3034:

bq. so i think its not an incompatible change. Please provide your opinion on 
the same.

Sorry, I missed that piece.

bq. IIUC SystemMetricsPublisher.publish*Event methods can determine which 
version of ATS to publish and can post it accordingly ?

I meant in the current approach SystemMetricsPublisher can be self contained. 
RMTimelineCollector can be a private stuff in SystemMetricsPublisher, 
constructed and started there. It's not necessary to be visible in RM and its 
context objects. 

bq. we might not require much of the functionality of SystemMetricsPublisher 
and it will be just delegating the calls to RMTimelineCollector.

I'm not sure about if there's previous discussion about the way for RM to put 
entities, but this approach sound cleaner, and in this approach, I don't think 
we should couple RMTimelineCollector and SystemMetricsPublisher. Keeping 
SystemMetricsPublisher separate, we can easily deprecate and even remove it 
from the code base later. Moreover, we can keep the existing config as what it 
is now, and create a new config to control starting v2 RM writing data stack.

> [Collector wireup] Implement RM starting its timeline collector
> ---------------------------------------------------------------
>                 Key: YARN-3034
>                 URL: https://issues.apache.org/jira/browse/YARN-3034
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: timelineserver
>            Reporter: Sangjin Lee
>            Assignee: Naganarasimha G R
>         Attachments: YARN-3034-20150312-1.patch, YARN-3034.20150205-1.patch, 
> YARN-3034.20150316-1.patch, YARN-3034.20150318-1.patch, 
> YARN-3034.20150320-1.patch
> Per design in YARN-2928, implement resource managers starting their own ATS 
> writers.

This message was sent by Atlassian JIRA

Reply via email to