Mailing-List: contact yarn-issues-help@hadoop.apache.org; run by ezmlm
Precedence: bulk
Date: Mon, 25 Jul 2016 13:15:20 +0000 (UTC)
From: "Feng Yuan (JIRA)" <jira@apache.org>
To: yarn-issues@hadoop.apache.org
Message-ID: <JIRA.12992192.1469436479000.123078.1469452520607@Atlassian.JIRA>
In-Reply-To: <JIRA.12992192.1469436479000@Atlassian.JIRA>
References: <JIRA.12992192.1469436479000@Atlassian.JIRA> <JIRA.12992192.1469436479526@arcas>
Subject: [jira] [Updated] (YARN-5427) Labeled apps cant use
 <DEFAULT_PARTITION> resources in 2.8
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable
archived-at: Mon, 25 Jul 2016 13:15:22 -0000


     [ https://issues.apache.org/jira/browse/YARN-5427?page=3Dcom.atlassian=
.jira.plugin.system.issuetabpanels:all-tabpanel ]

Feng Yuan updated YARN-5427:
----------------------------
    Attachment: QQ=E6=88=AA=E5=9B=BE20160725205642.png
                capacity-scheduler.xml

Hi all above is my configs,i can run a job which can get containers from pa=
rtition x and partition default with this configs in hdp 2.7.2.
I setup two nodes:
                  node "196" have 6GB with label "default".
                  node "197" have 6GB with label "x".
                  in my configs,queue A have 100% resource of them two.
                  and queue A`s default-node-label-expression is "x".(defau=
lt empty dont work like below.)
                  When i submit job in queue A,containers(one allcate 2GB) =
spread over 196 and 197 as the picture show.
                  I think this  phenomenon  contrarys to what [~Naganarasim=
ha]     says.
                 And i cant run this scene in 2.8.
                 How could i explain this?

> Labeled apps cant use <DEFAULT_PARTITION> resources in 2.8
> ----------------------------------------------------------
>
>                 Key: YARN-5427
>                 URL: https://issues.apache.org/jira/browse/YARN-5427
>             Project: Hadoop YARN
>          Issue Type: Bug
>          Components: capacity scheduler
>    Affects Versions: 2.8.0
>            Reporter: Feng Yuan
>             Fix For: 2.8.0
>
>         Attachments: QQ=E6=88=AA=E5=9B=BE20160725205642.png, capacity-sch=
eduler.xml
>
>
> Queue A use partition X,When x is used up,app cant get resource from <DEF=
AULT_PARTITION>.
> I find this behavior is different from 2.7.x.Also i find this in the docu=
ment of 2.7.2 as blew:
> "All queues can access to nodes without label"--https://hadoop.apache.org=
/docs/stable/hadoop-yarn/hadoop-yarn-site/NodeLabel.html
> I think it means if queue A`s partition x is used up,app can get containe=
rs from DEFAULT_PARTITION? Do i misstake?
> And I can confirm my configs is correct.Because i can run profectlly in 2=
.7.x use this configation too.
> I have observed this in 2.8: "<DEFAULT_PARTITION> is marked as <ECLUSIVE_=
PARTITION>" in webui.Is this the reason?=20


--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: yarn-issues-help@hadoop.apache.org