Wangda Tan commented on YARN-4003:

It's a very good point. I would prefer let AM start if there's any idle 
resource, I think we can use parent's available-capacity to compute 
AM-resource-limit, and preempt if other queues ask resource.

Currently in ProportionalCPP, it will preempt AM container only if there's no 
other container to preempt, I feel it may not work properly if we compute 
AM-resource-limit according to parent's available-capacity. Considering a case 
that there're lots of AM launched in a queue with small guaranteed capacity, it 
can end up with a queue's resources are all exhausted by AM containers.


> ReservationQueue inherit getAMResourceLimit() from LeafQueue, but behavior is 
> not consistent
> --------------------------------------------------------------------------------------------
>                 Key: YARN-4003
>                 URL: https://issues.apache.org/jira/browse/YARN-4003
>             Project: Hadoop YARN
>          Issue Type: Bug
>          Components: resourcemanager
>            Reporter: Carlo Curino
>         Attachments: YARN-4003.patch
> The inherited behavior from LeafQueue (limit AM % based on capacity) is not a 
> good fit for ReservationQueue (that have highly dynamic capacity). 

This message was sent by Atlassian JIRA

Reply via email to