Mailing-List: contact dev-help@hive.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@hive.apache.org
Date: Thu, 22 May 2014 23:47:01 +0000 (UTC)
From: "Swarnim Kulkarni (JIRA)" <jira@apache.org>
To: hive-dev@hadoop.apache.org
Message-ID: <JIRA.12716215.1400797249085.5178.1400802421968@arcas>
In-Reply-To: <JIRA.12716215.1400797249085@arcas>
References: <JIRA.12716215.1400797249085@arcas>
Subject: [jira] [Commented] (HIVE-7116) HDFS FileSystem object cache causes
 permission issues in creating tmp directories
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit


    [ https://issues.apache.org/jira/browse/HIVE-7116?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14006623#comment-14006623 ] 

Swarnim Kulkarni commented on HIVE-7116:
----------------------------------------

Hi Vikram,

Couple of smaller comments on the patch.

{code}
LOG.info("Create dirs " + mkdirPath + " with permission " + fsPermission + " recursive " + recursive);
{code}

You probably don't need to manually concatenate the string but let log4j do it for you.

{code}
LOG.info("Create dirs {} with permission {} recursive {}", mkdirPath, fsPermission, recursive);
{code}

Secondly, I think that if you create a FileSystem using the newInstance method, it should be closed to avoid any resource leaks. Previosuly with the cached FS instance, the API was automatically handling that for you.

> HDFS FileSystem object cache causes permission issues in creating tmp directories
> ---------------------------------------------------------------------------------
>
>                 Key: HIVE-7116
>                 URL: https://issues.apache.org/jira/browse/HIVE-7116
>             Project: Hive
>          Issue Type: Bug
>          Components: HiveServer2, Tez
>    Affects Versions: 0.13.0
>            Reporter: Vikram Dixit K
>            Assignee: Vikram Dixit K
>         Attachments: HIVE-7116.1.patch
>
>
> We change permissions of the directory creation to 777 for HiveServer 2 operation and it turns out that because of HDFS caching, it does not reflect once created. We need to use the non-cached version of the API to get a FileSystem object to fix this.


--
This message was sent by Atlassian JIRA
(v6.2#6252)