[
https://issues.apache.org/jira/browse/YARN-3047?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14366917#comment-14366917
]
Varun Saxena commented on YARN-3047:
------------------------------------
Thanks a lot [~zjshen] for the review.
bq. 1. No need to change timeline/TimelineEvents.java.
Ok.
bq. 2. In YarnConfiguration, how about we still reusing the existing timeline
service config. I propose config reuse because there doesn't exist the use case
that we start old timeline server and the new timeline reader server together.
And change in WebAppUtils should be not necessary too.
Same config has been used by aggregator as well. Thats why kept a new config. I
guess it is possible that reader runs on the same node as aggregator
bq. 3. NameValuePair is for internal usage only. Let's keep it in the timeline
service module?
Its in timeline service package itself i.e.
{{hadoop-yarn-server-timelineservice/src/main/java/org/apache/hadoop/yarn/server/timelineservice/NameValuePair.java}}.
Did you mean something else ?
bq. Rename TimelineReaderStore to TimelineReader.
Ok.
bq. I think we don't need to have NullTimelineReader. Instead, we should have a
POC implementation based on local FS like FileSystemTimelineWriterImpl. But we
can defer this work in a separate jira if the implementation is not
straightforward.
Yes NullTimelineReader was just to compile the code as TimelineReader store
would be an interface. Plan to have FS based implementation as part of
YARN-3051. Will update a patch for it once this goes in. Probably store related
code can be removed from this JIRA and handled completely as part of YARN-3051
to have a focussed review. Thoughts ?
bq. 5. TimelineReaderServer -> TimelineWebServer? For
startTimelineReaderWebApp, can we do something similar to
TimelineAggregatorsCollection#startWebApp.
The intention for TimelineReaderServer was not to have it merely act as a REST
endpoint. Hence not the name TimelineWebServer. TimelineReaderServer would use
RPC as well for instance to serve request coming from YARN CLI. Commands such
as "yarn application" used to contact AHS if app was not found in RM. This
should now be handled by Timeline Reader. For this, I plan to raise another
JIRA, once this one goes in.
bq. 6. Add the command in yarn and yarn.cmd to start the server.
This as per discussion with Sangjin will be done as part of YARN-3048.
I will probably update a document regarding TimelineReader as soon as possible.
> [Data Serving] Set up ATS reader with basic request serving structure and
> lifecycle
> -----------------------------------------------------------------------------------
>
> Key: YARN-3047
> URL: https://issues.apache.org/jira/browse/YARN-3047
> Project: Hadoop YARN
> Issue Type: Sub-task
> Components: timelineserver
> Reporter: Sangjin Lee
> Assignee: Varun Saxena
> Attachments: YARN-3047.001.patch, YARN-3047.02.patch
>
>
> Per design in YARN-2938, set up the ATS reader as a service and implement the
> basic structure as a service. It includes lifecycle management, request
> serving, and so on.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
