Mailing-List: contact yarn-issues-help@hadoop.apache.org; run by ezmlm
Precedence: bulk
Date: Tue, 4 Apr 2017 07:46:41 +0000 (UTC)
From: "Masatake Iwasaki (JIRA)" <jira@apache.org>
To: yarn-issues@hadoop.apache.org
Message-ID: <JIRA.13013054.1476755224000.198330.1491292001920@Atlassian.JIRA>
In-Reply-To: <JIRA.13013054.1476755224000@Atlassian.JIRA>
References: <JIRA.13013054.1476755224000@Atlassian.JIRA> <JIRA.13013054.1476755224816@jira-lw-us.apache.org>
Subject: [jira] [Updated] (YARN-5748) Backport YARN-5718 to branch-2
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit
archived-at: Tue, 04 Apr 2017 07:46:46 -0000


     [ https://issues.apache.org/jira/browse/YARN-5748?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Masatake Iwasaki updated YARN-5748:
-----------------------------------
    Attachment: YARN-5748-branch-2.001.patch

I saw this issue recently. I attached a patch for branch-2 going with option #2. The 001 adds on/off switch to retry policy while keeping default behaviour. It also adds regression test using FileSystemRMStateStore. I think it is ok to change the default value from true to false here or later if we have confidence.

> Backport YARN-5718 to branch-2
> ------------------------------
>
>                 Key: YARN-5748
>                 URL: https://issues.apache.org/jira/browse/YARN-5748
>             Project: Hadoop YARN
>          Issue Type: Bug
>            Reporter: Junping Du
>            Assignee: Junping Du
>         Attachments: YARN-5748-branch-2.001.patch
>
>
> In YARN-5718, we have identify several unnecessary config to over-write HDFS client behavior in several components of YARN (FSRMStore, TimelineClient, NodeLabelStore, etc.) which cause job failure in some cases (NN HA, etc.) - that's definitely belongs to bug. In YARN-5718, we proposed to remove the config as it shouldn't be supposed to work, which get committed to trunk already as alpha stage has more flexibility for incompatible changes. In branch-2, we want to play a bit more safe and get more discussion. 
> Obviously, there are several options here:
> 1. Don't fix anything, let bug exist
> 2. Fix the bug, but keep the configuration, or mark it deprecated and add some explanation to say this configuration is not supposed to work any more.
> 3. Exactly like YARN-5718, fix the bug and remove the unnecessary configuration.
> This ticket is filed for more discussion.


--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

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