Li Lu commented on YARN-3034:

Hi [~Naganarasimha], thanks for the patch! I briefly looked at it, and have 
some questions about it. 

* A general question, I think there are some inconsistencies between this patch 
and the proposed solution for aggregators. In the original design, it is 
proposed that we need to organize application level aggregators into 
collections (either on the NMs or on the RM, supposedly implemented as 
AppLevelServiceManager? ), and the servers launches its own collection. I could 
not find related logic in this patch, am I missing anything here? 

* I noticed that you refactored some metrics related code in the RM, moving 
part them into the new RMTimelineAggregator. Maybe in this JIRA we'd like to 
focus on setting up the "wiring" for the aggregator (collections) on the RM, 
rather than moving into the details of the timeline data? We can always resolve 
those problems in a separate JIRA after we set up the base infrastructure for 
timeline v2. 

* About source code organization: currently you're putting RMTimelineAggregator 
into the hadoop-yarn-server-resourcemanager module, under the package 
org.apache.hadoop.yarn.server.resourcemanager.metrics package. I'm not sure if 
that's a place we'd like it to be in. YARN-3166 keeps track of code 
organization related discussions, and you're more than welcome to join the 
discussion there. 

I think for now in this JIRA, maybe we want to firstly focus on making RM 
launches its aggregator collection (not blocked by any other JIRAs, may 
interfere with aggregator refactoring)?

> [Aggregator wireup] Implement RM starting its ATS writer
> --------------------------------------------------------
>                 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-3034.20150205-1.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