Mailing-List: contact user-help@cassandra.apache.org; run by ezmlm
Precedence: bulk
Reply-To: user@cassandra.apache.org
Received-SPF: pass (athena.apache.org: domain of edlinuxguru@gmail.com
 designates 209.85.223.172 as permitted sender)
MIME-Version: 1.0
In-Reply-To: 
 <E8EE4043A42D0846AD3CC043A0B6A84A06559EF3@NARSIL.us.int.genesyslab.com>
References: 
 <E8EE4043A42D0846AD3CC043A0B6A84A06559EF3@NARSIL.us.int.genesyslab.com>
Date: Thu, 15 Nov 2012 19:29:58 -0500
Message-ID: 
 <CAENxBwxt6WC66swiUk6in-kbQRYw=_inF42+KxVs7nzSKwzzAQ@mail.gmail.com>
Subject: Re: Question regarding the need to run nodetool repair
From: Edward Capriolo <edlinuxguru@gmail.com>
To: "user@cassandra.apache.org" <user@cassandra.apache.org>
Content-Type: multipart/alternative; boundary=bcaec51b1d6918316004ce91de3d

--bcaec51b1d6918316004ce91de3d
Content-Type: text/plain; charset=ISO-8859-1

On Thursday, November 15, 2012, Dwight Smith <Dwight.Smith@genesyslab.com>
wrote:
> I have a 4 node cluster,  version 1.1.2, replication factor of 4,
read/write consistency of 3, level compaction. Several questions.
>
>
>
> 1)      Should nodetool repair be run regularly to assure it has
completed before gc_grace?  If it is not run, what are the exposures?

Yes. Lost tombstones could cause deleted data to re appear.
>
> 2)      If a node goes down, and is brought back up prior to the 1 hour
hinted handoff expiration, should repair be run immediately?

If node is brought up prior to 1 hour. You should let the hints replay.
 Repair is always safe to run.
>
> 3)      If the hinted handoff has expired, the plan is to remove the node
and start a fresh node in its place.  Does this approach cause problems?
>
You only need to join a fresh mode if the node was down longer then gc
grace. Default is 10 days.
>
>
> Thanks
>
>

If you read and write at quorum and run repair regularly you can worry less
about the things above because they are essentially non factors.

--bcaec51b1d6918316004ce91de3d
Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable

<br><br>On Thursday, November 15, 2012, Dwight Smith &lt;<a href=3D"mailto:=
Dwight.Smith@genesyslab.com">Dwight.Smith@genesyslab.com</a>&gt; wrote:<br>=
&gt; I have a 4 node cluster, =A0version 1.1.2, replication factor of 4, re=
ad/write consistency of 3, level compaction. Several questions.<br>
&gt;<br>&gt; =A0<br>&gt;<br>&gt; 1)=A0=A0=A0=A0=A0 Should nodetool repair b=
e run regularly to assure it has completed before gc_grace?=A0 If it is not=
 run, what are the exposures?<br><br>Yes. Lost tombstones could cause delet=
ed data to re appear.<br>
&gt;<br>&gt; 2)=A0=A0=A0=A0=A0 If a node goes down, and is brought back up =
prior to the 1 hour hinted handoff expiration, should repair be run immedia=
tely?<br><br>If node is brought up prior to 1 hour. You should let the hint=
s replay. =A0Repair is always safe to run.<br>
&gt;<br>&gt; 3)=A0=A0=A0=A0=A0 If the hinted handoff has expired, the plan =
is to remove the node and start a fresh node in its place.=A0 Does this app=
roach cause problems?<br>&gt;<br>You only need to join a fresh mode if the =
node was down longer then gc grace. Default is 10 days.<br>
&gt; =A0<br>&gt;<br>&gt; Thanks<br>&gt;<br>&gt;<br><br>If you read and writ=
e at quorum and run repair regularly you can worry less about the things ab=
ove because they are essentially non factors.

--bcaec51b1d6918316004ce91de3d--