[
https://issues.apache.org/jira/browse/YARN-5325?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15537102#comment-15537102
]
Carlo Curino commented on YARN-5325:
------------------------------------
[~subru] Thanks for the review, I will start addressing it right away. Few
comments answers:
Regarding "Use Math.round instead of (int) Math.ceil in splitIndividualAny." I
don't think we can, as this might lead to less than what the user asked, e.g.,
I have 3 subclusters with equal weights/load and you ask for 1 container
anywhere... I will end up rounding down the 0.333 containers for each
subclusters and give you zero containers back.
testStressTest was mostly to measure how much it takes to run a large number of
big allocations. Do you think we should measure time and assert it should be <
X? (Seems bad, though the timing information are generally useful as we
develop/evolve this). I guess as it stands it only checks that large/many
allocations don't end up throwing off anything. I am unsure about this.
Re: thread safety I am assuming that, though we should validate it carefully...
I am happy to walk over the getHeadroomWeighting stuff with you... I am also
mostly convinced is correct, but must be validated well.
Thanks,
Carlo
> Stateless ARMRMProxy policies implementation
> --------------------------------------------
>
> Key: YARN-5325
> URL: https://issues.apache.org/jira/browse/YARN-5325
> Project: Hadoop YARN
> Issue Type: Sub-task
> Components: nodemanager, resourcemanager
> Affects Versions: YARN-2915
> Reporter: Carlo Curino
> Assignee: Carlo Curino
> Attachments: YARN-5325-YARN-2915.05.patch,
> YARN-5325-YARN-2915.06.patch, YARN-5325-YARN-2915.07.patch,
> YARN-5325.01.patch, YARN-5325.02.patch, YARN-5325.03.patch, YARN-5325.04.patch
>
>
> This JIRA tracks policies in the AMRMProxy that decide how to forward
> ResourceRequests, without maintaining substantial state across decissions
> (e.g., broadcast).
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
