Karthik Kambatla commented on YARN-1012:

bq. The bigger reason is vcore is more of a construct created for scheduling. 
We can have a vcore-millis which represents more of the resources reserved as 
opposed to real utilization. Real utilization can be cpu-wall-time, avg cpu 
utilization etc.

I see your point. That said, NM is sending information across to RM solely 
because the scheduler can take it into consideration. It would definitely be 
simpler if the scheduler doesn't have to convert "real" utilization to vcores. 
NodeManagers are better places to do that. I would like for the heartbeat 
itself to have the utilization information in rounded-up vcores on 
milli-vcores. Should we store both? 

> NM should report resource utilization of running containers to RM in heartbeat
> ------------------------------------------------------------------------------
>                 Key: YARN-1012
>                 URL: https://issues.apache.org/jira/browse/YARN-1012
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: nodemanager
>    Affects Versions: 2.7.0
>            Reporter: Arun C Murthy
>            Assignee: Inigo Goiri
>         Attachments: YARN-1012-1.patch, YARN-1012-2.patch, YARN-1012-3.patch, 
> YARN-1012-4.patch, YARN-1012-5.patch, YARN-1012-6.patch, YARN-1012-7.patch

This message was sent by Atlassian JIRA

Reply via email to