[
https://issues.apache.org/jira/browse/YARN-6428?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16069011#comment-16069011
]
Sunil G commented on YARN-6428:
-------------------------------
Thanks [[email protected]] and [~bibinchundatt] for the
clarifications.
If we use a very big value in long, then it could cause an overflow once we
multiple with 10^6. We moved from int to long to support -ve round of cases, so
I am also not much lenient in putting a loop hole there. Instead could we use
BigDecimal to set precision point. I could see that we are using BigInteger in
few places, so it may be fine. I will wait for [~leftnoteasy] here.
My alternate proposal is something like this
{code}
BigDecimal bd = new BigDecimal(by).setScale(2,
RoundingMode.HALF_EVEN);
double by = bd.doubleValue();
{code}
I will wait for more comments too at this point.
> Queue AM limit is not honored in CS always
> -------------------------------------------
>
> Key: YARN-6428
> URL: https://issues.apache.org/jira/browse/YARN-6428
> Project: Hadoop YARN
> Issue Type: Bug
> Reporter: Bibin A Chundatt
> Assignee: Bibin A Chundatt
> Attachments: YARN-6428.0001.patch, YARN-6428.0002.patch
>
>
> Steps to reproduce
> ------------------------
> Setup cluster with 40 GB and 40 vcores with 4 Node managers with 10 GB each.
> Configure 100% to default queue as capacity and max am limit as 10 %
> Minimum scheduler memory and vcore as 512,1
> *Expected*
> AM limit 4096 and 4 vores
> *Actual*
> AM limit 4096+512 and 4+1 vcore
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
