Carlo Curino commented on YARN-4003:

I think you are correct, but given that the size of the allocation for a 
ReservationQueue can fluctuate substantially, I think we need to rely on either 
hard-limits on the number of applications running per queue, or careful user 
submissions will need to protect from the bad scenario you describe. 
Picking the opposite strategy (using actual reservation size to limit number of 
AMs) prevents us to scavange resources early in the reservation, and I am not 
sure even works properly if the reservation "shrinks" (does ProportionalCPP 
kill AMs running in a queue, if the capacity is shrinked?).

I agree with your concerns, but I don't see a better way out.

> 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