Mailing-List: contact yarn-issues-help@hadoop.apache.org; run by ezmlm
Precedence: bulk
Date: Wed, 3 Jan 2018 06:33:00 +0000 (UTC)
From: "Konstantinos Karanasos (JIRA)" <jira@apache.org>
To: yarn-issues@hadoop.apache.org
Message-ID: <JIRA.13127122.1514228156000.561394.1514961180659@Atlassian.JIRA>
In-Reply-To: <JIRA.13127122.1514228156000@Atlassian.JIRA>
References: <JIRA.13127122.1514228156000@Atlassian.JIRA> <JIRA.13127122.1514228156783@jira-lw-us.apache.org>
Subject: [jira] [Commented] (YARN-7682) Expose canAssign method in the
 PlacementConstraintManager
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable
archived-at: Wed, 03 Jan 2018 06:33:06 -0000


    [ https://issues.apache.org/jira/browse/YARN-7682?page=3Dcom.atlassian.=
jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=3D16309=
205#comment-16309205 ]=20

Konstantinos Karanasos commented on YARN-7682:
----------------------------------------------

Thanks for the patch, [~pgaref]. Two main comments:
* I think that the functions you push in the getNodeCardinalityByOp should =
be reversed. That is, I think when you have multiple tags, you should look =
at the highest cardinality one in the node/rack, and your min cardinality o=
f the constraint should be below that. Same for the max cardinality of the =
constraint (should be above the min of the actual cardinalities of the give=
n tags). So essentially you would define your minScopeCardinality using the=
 Long::max, and your maxScopeCardinality using the Long::min. Then everythi=
ng goes as is. You can also make a check that your minScopeCardinality <=3D=
 maxScopeCardinality, just to be on the safe side. Does it make sense?
* Do we need the line right after the comment =E2=80=9C// Make sure Anti-af=
finity satisfies hard upper limit=E2=80=9D?

Nits:
* Let=E2=80=99s fill out the javadoc comments in canSatisfyConstraints
* In canSatisfySingleConstraintExpression the javadoc is a bit ambiguous. "=
The node or rack should satisfy the constraints that are enabled by the giv=
en allocation tags". Also let=E2=80=99s add javadoc comments for parameters=
 its parameters (I know it is private, but it helps).
* In both methods you mention Node in the javadoc but you actually support =
RACK too.

If the above are fixed, +1 from me.

> Expose canAssign method in the PlacementConstraintManager
> ---------------------------------------------------------
>
>                 Key: YARN-7682
>                 URL: https://issues.apache.org/jira/browse/YARN-7682
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>            Reporter: Arun Suresh
>            Assignee: Panagiotis Garefalakis
>         Attachments: YARN-7682-YARN-6592.001.patch, YARN-7682-YARN-6592.0=
02.patch, YARN-7682-YARN-6592.003.patch, YARN-7682.wip.patch
>
>
> As per discussion in YARN-7613. Lets expose {{canAssign}} method in the P=
lacementConstraintManager that takes a sourceTags, applicationId, Scheduler=
Node and AllocationTagsManager and returns true if constraints are not viol=
ated by placing the container on the node.
> I prefer not passing in the SchedulingRequest, since it can have > 1 numA=
llocations. We want this api to be called for single allocations.


--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

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