Mailing-List: contact user-help@hadoop.apache.org; run by ezmlm
Precedence: bulk
Reply-To: user@hadoop.apache.org
Received-SPF: pass (athena.apache.org: domain of harsh@cloudera.com designates
 209.85.210.174 as permitted sender)
MIME-Version: 1.0
In-Reply-To: 
 <CADokdcM+UJkswn6BGr+kpOr9XrS7ntEsAUHu9Y2zTkN_oOU3=g@mail.gmail.com>
References: <BLU171-W96F6B3C21FA1040D7CCB282D40@phx.gbl>
 <CACvhJWc=o6a-qdjaaX0fbOwhaKg8gmfTesYK_V_xASQCsySTgg@mail.gmail.com>
 <CALr1C9pvU8p7BKgm0eZZfwpDFbEXe4km94bC2FCOKbgJox=_9Q@mail.gmail.com>
 <CAOcnVr0VsWXkB6m-tpXejOMwsdqS2D3TFDW2_9qSwGZLP6L6pQ@mail.gmail.com>
 <CADokdcM+UJkswn6BGr+kpOr9XrS7ntEsAUHu9Y2zTkN_oOU3=g@mail.gmail.com>
From: Harsh J <harsh@cloudera.com>
Date: Sat, 23 Mar 2013 11:12:53 +0530
Message-ID: 
 <CAOcnVr1qbwoZg63bWxZ=OneTR_Lrz8ZS1bpjStqQN1oacVeGLw@mail.gmail.com>
Subject: Re: Cluster lost IP addresses
To: "<user@hadoop.apache.org>" <user@hadoop.apache.org>,
 Chris Embree <chris@embree.us>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable

Hi Chris,

Where exactly are you seeing issues with change of NN/DN IPs? I've
never encountered trouble on IP changes (I keep moving across networks
everyday and the HDFS plus MR I run both stand tall without requiring
a restart). We do not store (generally) nor rely on IP addresses. An
exclusion may apply to files under construction I think, but a
properly shutdown cluster pre-move would not have that and such files
wouldn't matter too much in such a scenario anyway.

Obviously, a hostname change could cause issues.

FWIW, you can easily take any person's fsimage from across the world
and start your NN on top of that and add in new DNs with the block
data under them and setup the HDFS cluster. This is rather painless
and well-built and goes to show that its not really IP dependent in
any way. Please do elaborate.

On Sat, Mar 23, 2013 at 10:52 AM, Chris Embree <cembree@gmail.com> wrote:
> Hey John,
>
> Make sure your /etc/hosts ( or DNS) is up to date and  any topology scrip=
ts
> are updated.  Unfortunately, NN is pretty dumb about IP's vs. Hostnames.
>
> BTW, NN devs.  Seriously?  You rely on IP addr instead of hostname?  Some=
one
> should probably be shot or at least be responsible for fixing this
> abomination. Sad that this code was released GA.
>
> Sorry folks.  HDFS/Mapred is really cool tech, I'm just jaded about this
> kind of silliness.
>
> In my Not So Humble Opinion.
> Chris
>
>
> On Sat, Mar 23, 2013 at 1:12 AM, Harsh J <harsh@cloudera.com> wrote:
>>
>> NameNode does not persist block locations; so this is still
>> recoverable if the configs are changed to use the new set of hostnames
>> to bind to/look up.
>>
>> On Sat, Mar 23, 2013 at 9:01 AM, Azuryy Yu <azuryyyu@gmail.com> wrote:
>> > it has issues, namenode save blockid->nodes, using ip addr if your
>> > slaves
>> > config file using ip addr instead of hostname.
>> >
>> > On Mar 23, 2013 10:14 AM, "Balaji Narayanan (=E0=AE=AA=E0=AE=BE=E0=AE=
=B2=E0=AE=BE=E0=AE=9C=E0=AE=BF =E0=AE=A8=E0=AE=BE=E0=AE=B0=E0=AE=BE=E0=AE=
=AF=E0=AE=A3=E0=AE=A9=E0=AF=8D)"
>> > <lists@balajin.net> wrote:
>> >>
>> >> Assuming you are using hostnAmes and not ip address in your config
>> >> files
>> >> What happens when you start the cluster? If you are using IP address =
in
>> >> your
>> >> configs just update them and start. It should work with no issues.
>> >>
>> >> On Friday, March 22, 2013, John Meza wrote:
>> >>>
>> >>> I have a 18 node cluster that had to be physically moved.
>> >>> Unfortunately all the ip addresses were lost (recreated).
>> >>>
>> >>> This must have happened to someone before.
>> >>> Nothing else on the machines has been changed. Most importantly the
>> >>> data
>> >>> in HDFS is still sitting there.
>> >>>
>> >>> Is there a way to recover this cluster to a useable state?
>> >>> thanks
>> >>> John
>> >>
>> >>
>> >>
>> >> --
>> >> http://balajin.net/blog
>> >> http://flic.kr/balajijegan
>>
>>
>>
>> --
>> Harsh J
>
>


--=20
Harsh J