Sangjin Lee commented on YARN-3049:

I thought that the application created event would be written by the RM (and 
its embedded collector), no? So I'm not sure if the writer (for the app 
timeline collector) being bound to the session of the application is an issue. 
Maybe I misunderstood your comment?

In essence, the flow of control that I was thinking of is not really different 
than your v.3 patch. My point was more about the way we're passing that 
information. I think it should be possible from inside 
{{HBaseTimelineWriterImpl}} to detect that it received an application created 
event (likely originating from RM) and trigger writing to these tables.

Also, note that we want to store the application created timestamp, and also 
application finished event along with its timestamp. That's not for this table 
but for other tables that are mentioned in the schema proposal doc. To be able 
to do these as well, it would be most natural to do it based on seeing these 


> [Storage Implementation] Implement storage reader interface to fetch raw data 
> from HBase backend
> ------------------------------------------------------------------------------------------------
>                 Key: YARN-3049
>                 URL: https://issues.apache.org/jira/browse/YARN-3049
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: timelineserver
>            Reporter: Sangjin Lee
>            Assignee: Zhijie Shen
>         Attachments: YARN-3049-WIP.1.patch, YARN-3049-WIP.2.patch, 
> YARN-3049-WIP.3.patch, YARN-3049-YARN-2928.2.patch, 
> YARN-3049-YARN-2928.3.patch, YARN-3049-YARN-2928.4.patch
> Implement existing ATS queries with the new ATS reader design.

This message was sent by Atlassian JIRA

Reply via email to