Mailing-List: contact yarn-issues-help@hadoop.apache.org; run by ezmlm
Precedence: bulk
Date: Tue, 3 Jan 2017 03:25:58 +0000 (UTC)
From: "Tao Yang (JIRA)" <jira@apache.org>
To: yarn-issues@hadoop.apache.org
Message-ID: <JIRA.13030678.1482838736000.625755.1483413958604@Atlassian.JIRA>
In-Reply-To: <JIRA.13030678.1482838736000@Atlassian.JIRA>
References: <JIRA.13030678.1482838736000@Atlassian.JIRA> <JIRA.13030678.1482838736517@arcas>
Subject: [jira] [Commented] (YARN-6029) CapacityScheduler deadlock when
 ParentQueue#getQueueUserAclInfo is called by Thread_A at the moment that
 Thread_B calls LeafQueue#assignContainers to release a reserved container
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable
archived-at: Tue, 03 Jan 2017 03:26:00 -0000


    [ https://issues.apache.org/jira/browse/YARN-6029?page=3Dcom.atlassian.=
jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=3D15794=
011#comment-15794011 ]=20

Tao Yang commented on YARN-6029:
--------------------------------

Hi=EF=BC=8C[~Naganarasimha]
I thought the application could be finished in another thread which is call=
ing LQ#finishApplicationAttempt meanwhile. Perhaps app will still assign co=
ntainer after finished.

> CapacityScheduler deadlock when ParentQueue#getQueueUserAclInfo is called=
 by Thread_A at the moment that Thread_B calls LeafQueue#assignContainers t=
o release a reserved container
> -------------------------------------------------------------------------=
---------------------------------------------------------------------------=
------------------------------
>
>                 Key: YARN-6029
>                 URL: https://issues.apache.org/jira/browse/YARN-6029
>             Project: Hadoop YARN
>          Issue Type: Bug
>          Components: capacityscheduler
>    Affects Versions: 2.8.0
>            Reporter: Tao Yang
>            Assignee: Wangda Tan
>            Priority: Critical
>         Attachments: YARN-6029-branch-2.8.001.patch, YARN-6029.001.patch,=
 YARN-6029.002.patch, deadlock.jstack
>
>
> When ParentQueue#getQueueUserAclInfo is called (e.g. a client calls YarnC=
lient#getQueueAclsInfo) just at the moment that LeafQueue#assignContainers =
is called and before notifying parent queue to release resource (should rel=
ease a reserved container), then ResourceManager can deadlock. I found this=
 problem on our testing environment for hadoop2.8.
> Reproduce the deadlock in chronological order
> * 1. Thread A (ResourceManager Event Processor) calls synchronized LeafQu=
eue#assignContainers (got LeafQueue instance lock of queue root.a)
> * 2. Thread B (IPC Server handler) calls synchronized ParentQueue#getQueu=
eUserAclInfo (got ParentQueue instance lock of queue root), iterates over c=
hildren queue acls and is blocked when calling synchronized LeafQueue#getQu=
eueUserAclInfo (the LeafQueue instance lock of queue root.a is hold by Thre=
ad A)
> * 3. Thread A wants to inform the parent queue that a container is being =
completed and is blocked when invoking synchronized ParentQueue#internalRel=
easeResource method (the ParentQueue instance lock of queue root is hold by=
 Thread B)
> I think the synchronized modifier of LeafQueue#getQueueUserAclInfo can be=
 removed to solve this problem, since this method appears to not affect fie=
lds of LeafQueue instance.
> Attach patch with UT for review.


--
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