Mailing-List: contact hdfs-issues-help@hadoop.apache.org; run by ezmlm
Precedence: bulk
Reply-To: hdfs-issues@hadoop.apache.org
Date: Wed, 3 Oct 2012 07:37:08 +1100 (NCT)
From: "Suresh Srinivas (JIRA)" <jira@apache.org>
To: hdfs-issues@hadoop.apache.org
Message-ID: <1689735299.156118.1349210228333.JavaMail.jiratomcat@arcas>
In-Reply-To: <658093001.59618.1347307568190.JavaMail.jiratomcat@arcas>
Subject: [jira] [Commented] (HDFS-3912) Detecting and avoiding stale
 datanodes for writing
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit


    [ https://issues.apache.org/jira/browse/HDFS-3912?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13468045#comment-13468045 ] 

Suresh Srinivas commented on HDFS-3912:
---------------------------------------

# Remove HeartbeatManager#checkStaleNodes and use DatanodeManager#checkStaleNodes instead
# What happens when ratio is configured invalid?
# when calculating the ration in HeatbeatManager, you are accessing datanodes.size() outside synchronization block.
# Can we introduce a method in FSClusterStats to provide the cluster state of whether it is avoiding writes to stale nodes and avoid having to add DatanodeManager into BlockPlacementPolicy. This way, customer placemet policy implementations are not affected.
# I think we should create a separte jira to move some relevant methods such as getLiveNodes, stale nodes etc into DatanodeStatics interface.
# We should also add metrics related to stale datanodes.

                
> Detecting and avoiding stale datanodes for writing
> --------------------------------------------------
>
>                 Key: HDFS-3912
>                 URL: https://issues.apache.org/jira/browse/HDFS-3912
>             Project: Hadoop HDFS
>          Issue Type: Sub-task
>            Reporter: Jing Zhao
>            Assignee: nkeywal
>         Attachments: HDFS-3912.001.patch, HDFS-3912.002.patch, HDFS-3912.003.patch, HDFS-3912.004.patch
>
>
> 1. Make stale timeout adaptive to the number of nodes marked stale in the cluster.
> 2. Consider having a separate configuration for write skipping the stale nodes.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira