Junping Du commented on YARN-3039:

Thanks [~zjshen] for clarification here!
bq.  It sounds reasonable to have aggregator as the client and NM as the 
server. However, my major point is not about which side is the server/client, 
but about the rpc call to trigger app-level aggregator initialization.
+1. That's good point we all agree I think. 

bq. So in this work, since we have involved the rpc call, I raise the thought 
of using the rpc call to start app-level aggregator instead of aux service 
lifecycle event handler. IMHO, it's necessary in the future when the aggregator 
no longer resides in the same process of NM, such that I can see the merit of 
uniforming the way to start app-level aggregator.
I like the idea here. However, The major work of this JIRA's patch now is on 
setting up a client/server protocol between aggregator and NM for service 
address registration and broadcast to AM and other NMs via RM. The work for 
starting of aggregator is covered by YARN-3030 (done for auxiliary service) and 
YARN-3033 (for standalone model). I would expect YARN-3033 could address the 
starting aggregator details given this patch is big enough. Thoughts?

> [Aggregator wireup] Implement ATS app-appgregator service discovery
> -------------------------------------------------------------------
>                 Key: YARN-3039
>                 URL: https://issues.apache.org/jira/browse/YARN-3039
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: timelineserver
>            Reporter: Sangjin Lee
>            Assignee: Junping Du
>         Attachments: Service Binding for applicationaggregator of ATS 
> (draft).pdf, Service Discovery For Application Aggregator of ATS (v2).pdf, 
> YARN-3039-no-test.patch, YARN-3039-v2-incomplete.patch, 
> YARN-3039-v3-core-changes-only.patch, YARN-3039-v4.patch, YARN-3039-v5.patch
> Per design in YARN-2928, implement ATS writer service discovery. This is 
> essential for off-node clients to send writes to the right ATS writer. This 
> should also handle the case of AM failures.

This message was sent by Atlassian JIRA

Reply via email to