Sunil G commented on YARN-2056:

+1 This make sense. I have a doubt though.

*max_ignored_over_capacity* will help to avoid the jitter when container sizes 
varies and some times we do a little more/less preemption from leaf queue than 
its defined capacity. So more or less its boil down to the resource size of 
containers. per-queue configuration for max_ignored_over_capacity will 
definitely give more control than now, but still if heterogeneous applications 
(in terms of container resource) keeps running same queue, it may be still hard 
to get a correct value.    

> Disable preemption at Queue level
> ---------------------------------
>                 Key: YARN-2056
>                 URL: https://issues.apache.org/jira/browse/YARN-2056
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: resourcemanager
>    Affects Versions: 2.4.0
>            Reporter: Mayank Bansal
>            Assignee: Eric Payne
> We need to be able to disable preemption at individual queue level

This message was sent by Atlassian JIRA

Reply via email to