Mailing-List: contact yarn-issues-help@hadoop.apache.org; run by ezmlm
Precedence: bulk
Reply-To: yarn-issues@hadoop.apache.org
Date: Mon, 10 Feb 2014 22:47:19 +0000 (UTC)
From: "Karthik Kambatla (JIRA)" <jira@apache.org>
To: yarn-issues@hadoop.apache.org
Message-ID: <JIRA.12691175.1390613902169.19871.1392072439817@arcas>
In-Reply-To: <JIRA.12691175.1390613902169@arcas>
References: <JIRA.12691175.1390613902169@arcas>
Subject: [jira] [Updated] (YARN-1641) ZK store - periodically create a dummy
 file to kick in fencing
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit


     [ https://issues.apache.org/jira/browse/YARN-1641?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Karthik Kambatla updated YARN-1641:
-----------------------------------

    Attachment: yarn-1641-1.patch

With this patch, the ZK-store periodically attempts writing a znode and transitions to standby if it fails to do so. 

The change to the test is precisely the effect we need - an RM shouldn't need to perform one of its operations that modify the store to realize it is not Active. 

> ZK store - periodically create a dummy file to kick in fencing
> --------------------------------------------------------------
>
>                 Key: YARN-1641
>                 URL: https://issues.apache.org/jira/browse/YARN-1641
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: resourcemanager
>    Affects Versions: 2.3.0
>            Reporter: Karthik Kambatla
>            Assignee: Karthik Kambatla
>         Attachments: yarn-1641-1.patch
>
>
> Fencing in ZK store kicks in when the RM tries to write something to the store. If the RM doesn't write anything to the store, it doesn't get fenced and can continue to assume being the Active. 
> By periodically writing a file (say, every RM_ZK_TIMEOUT_MS seconds), we can ensure it gets fenced.


--
This message was sent by Atlassian JIRA
(v6.1.5#6160)