Return-Path: Delivered-To: apmail-hadoop-hdfs-user-archive@minotaur.apache.org Received: (qmail 61181 invoked from network); 8 Oct 2010 06:31:32 -0000 Received: from unknown (HELO mail.apache.org) (140.211.11.3) by 140.211.11.9 with SMTP; 8 Oct 2010 06:31:32 -0000 Received: (qmail 94976 invoked by uid 500); 8 Oct 2010 06:31:32 -0000 Delivered-To: apmail-hadoop-hdfs-user-archive@hadoop.apache.org Received: (qmail 94730 invoked by uid 500); 8 Oct 2010 06:31:30 -0000 Mailing-List: contact hdfs-user-help@hadoop.apache.org; run by ezmlm Precedence: bulk List-Help: List-Unsubscribe: List-Post: List-Id: Reply-To: hdfs-user@hadoop.apache.org Delivered-To: mailing list hdfs-user@hadoop.apache.org Received: (qmail 94722 invoked by uid 99); 8 Oct 2010 06:31:29 -0000 Received: from athena.apache.org (HELO athena.apache.org) (140.211.11.136) by apache.org (qpsmtpd/0.29) with ESMTP; Fri, 08 Oct 2010 06:31:29 +0000 X-ASF-Spam-Status: No, hits=2.2 required=10.0 tests=HTML_MESSAGE,SPF_PASS X-Spam-Check-By: apache.org Received-SPF: pass (athena.apache.org: local policy) Received: from [128.105.6.20] (HELO sabe.cs.wisc.edu) (128.105.6.20) by apache.org (qpsmtpd/0.29) with ESMTP; Fri, 08 Oct 2010 06:31:24 +0000 Received: from xPC ([146.151.127.226]) (authenticated bits=0) by sabe.cs.wisc.edu (8.14.1/8.14.1) with ESMTP id o986V3UV021983 (version=TLSv1/SSLv3 cipher=AES128-SHA bits=128 verify=NO); Fri, 8 Oct 2010 01:31:03 -0500 Message-ID: <34124B4BB9DA4EA7B582AC2BE9C3F472@xPC> From: "Ozcan ILIKHAN" To: Cc: "Nick LeRoy" References: <0C88594C80CC403A88CBE0FDDD2AF64B@xPC> In-Reply-To: Subject: Re: Backup NameNode Date: Fri, 8 Oct 2010 01:30:56 -0500 MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_NextPart_000_0175_01CB6688.747E22C0" X-Priority: 3 X-MSMail-Priority: Normal Importance: Normal X-Mailer: Microsoft Windows Live Mail 14.0.8117.416 X-MIMEOLE: Produced By Microsoft MimeOLE V14.0.8117.416 This is a multi-part message in MIME format. ------=_NextPart_000_0175_01CB6688.747E22C0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable How should configuration files of Active and Backup NameNode should = differ from each other? Assume that our Active NameNode died, do we need to anything further DNS = record change? We start our Backup NameNode with "-backup" argument. = How to convert role to active? From: Dhruba Borthakur=20 Sent: Friday, October 08, 2010 12:49 AM To: hdfs-user@hadoop.apache.org=20 Subject: Re: Backup NameNode ActiveNameNode and BackupNamenode do not automatically exchange roles = ever. DataNodes are not informed. One can use Virtual IP failover or = DNS to switch so that DataNodes need not know that the Namenode has = moved from one machine to another.=20 hope this helps, dhruba On Wed, Oct 6, 2010 at 10:07 PM, Ozcan ILIKHAN = wrote: Hi all, Following questions are specific to v0.21.0. From the user guide it is not clear enough how Active NameNode and = Backup Namenode take over role from each other, how DateNode are = informed if active NameNode fails (or do they need to be informed?). = What is happening in disaster scnerios. While configuring cluster, assume 1 active NameNode, 1 Backup = NameNode, 1 Chekpoint NameNode, and the rest of the cluster is assumed = to be DataNodes. Which parameters for these 4 types of nodes' = hdfs-site.xml and core-site.xml should set, and how should they (config = files of 4 type of nodes ) differ from each other. By the way, in order to understand certain functionalities, most of = the time we need to go deep into source code. I think it will be good to = enrich Hadoop (especially HDFS) documentation. Thanks in advance, Ozcan. --=20 Connect to me at http://www.facebook.com/dhruba ------=_NextPart_000_0175_01CB6688.747E22C0 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable
How should configuration files of Active and = Backup=20 NameNode should differ from each other?
 
Assume that our Active NameNode died, do we = need to=20 anything further DNS record change? We start our Backup NameNode with = "-backup"=20 argument.  How to convert role to active?
 

Sent: Friday, October 08, 2010 12:49 AM
Subject: Re: Backup NameNode

ActiveNameNode and BackupNamenode do not automatically=20  exchange roles ever.  DataNodes are not informed. One can use = Virtual=20 IP failover  or DNS to switch so that DataNodes need not know that = the=20 Namenode has moved from one machine to another.=20

hope this helps,
dhruba

On Wed, Oct 6, 2010 at 10:07 PM, Ozcan ILIKHAN = <ilikhan@cs.wisc.edu> = wrote:
Hi all,
 
Following questions are specific to=20 v0.21.0.
 
From the user guide it is not clear = enough how=20 Active NameNode and Backup Namenode take over role from each other, = how=20 DateNode are informed if active NameNode  fails (or do = they=20 need to be informed?). What is happening in disaster = scnerios.
 
While configuring cluster,=20 assume 1 active NameNode, 1 Backup NameNode, 1 = Chekpoint=20 NameNode, and the rest of the cluster is assumed to be=20 DataNodes.  Which parameters for these 4 types of=20 nodes' hdfs-site.xml and core-site.xml  should set, and how = should=20 they (config files of 4 type of nodes ) differ from each=20 other.
 
By the way, in order to understand = certain=20 functionalities, most of the time we need to go deep into source code. = I think=20 it will be good to enrich Hadoop (especially=20 HDFS) documentation.
 
Thanks in advance,
Ozcan.


--
Connect to me at http://www.facebook.com/dhruba
------=_NextPart_000_0175_01CB6688.747E22C0--