Mailing-List: contact user-help@cassandra.apache.org; run by ezmlm
Precedence: bulk
Reply-To: user@cassandra.apache.org
Received-SPF: pass (nike.apache.org: domain of yulinyen@gmail.com designates
 74.125.83.44 as permitted sender)
MIME-Version: 1.0
In-Reply-To: 
 <CAKkz8Q0=58gOSkbJPuOoCxSNR1RqMXThb+Ta0XKAp_QA1Tx1jw@mail.gmail.com>
References: 
 <CAAaMp539dozHMBCuGAMRWEzWvGSmJRDS_j82fe07wJMNazz2TQ@mail.gmail.com>
	<CAKkz8Q0=58gOSkbJPuOoCxSNR1RqMXThb+Ta0XKAp_QA1Tx1jw@mail.gmail.com>
Date: Mon, 5 Dec 2011 17:47:34 +0800
Message-ID: 
 <CAAaMp516OAE_rS0+N=Bjb=Jox=H=h5ZEdx0owTgP2EFjQ2MXMQ@mail.gmail.com>
Subject: Re: DyanmicCompositeType bug?
From: Boris Yen <yulinyen@gmail.com>
To: user@cassandra.apache.org
Content-Type: multipart/alternative; boundary=0015174c18384a0dc004b35535a3

--0015174c18384a0dc004b35535a3
Content-Type: text/plain; charset=ISO-8859-1

It looks like using the dynamicComposite is a dangerous action. However,
one mistaken insertion causes the database to corrupt seems a little bit
too much. Is there anyway to just skip the wrong column.

On Mon, Dec 5, 2011 at 5:31 PM, Sylvain Lebresne <sylvain@datastax.com>wrote:

> On Mon, Dec 5, 2011 at 8:19 AM, Boris Yen <yulinyen@gmail.com> wrote:
> > Hi,
> >
> > I am using 0.8.7.
> >
> > I was trying to use DynamicComposite column. After I intentional added a
> > column (string:string:uuid) into a record which has previous columns
> > inserted with comparator (string:uuid:uuid). I got an exception on the
> > cassandra side and an timeout exception on the client (I was actually
> > expecting an validation exception on the client).
>
> By design, the DynamicComposite comparator cannot return a validation
> exception in that case. Or rather, since there is nothing specifying how
> a column should look like, the only way we could ever hope to return a
> validation exception would be to scan the entire row on insertion (which
> is out of question).
>
> > After that when I re-started cassandra, cassandra kept saying that
> > "comparator mismatch while comparing two dynamicCompositeType column..."
> and
> > cassandra never really got started.
> >
> > Is this a bug?
>
> No. As said above, this is something we cannot do anything about.
> The DynamicCompositeType comparator is by nature a very dangerous
> one. One column inserted in the wrong row can corrupt that row (with
> recovery involving losing data).
> In other words, don't use DynamicCompositeType. Don't use it unless
> you are absolutely sure you won't ever insert the wrong column in the
> wrong row *and* you absolutely need the DynamicCompositeType (by
> opposition to the faster, cheaper and safe CompositeType). And if you
> do, you have to assume the consequence if you do mess up.
>
> --
> Sylvain
>
> --
> Sylvain
>

--0015174c18384a0dc004b35535a3
Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable

It looks like using the dynamicComposite is a dangerous action. However, on=
e mistaken insertion causes the database to corrupt seems a little bit too =
much. Is there anyway to just skip the wrong column.<br><br><div class=3D"g=
mail_quote">
On Mon, Dec 5, 2011 at 5:31 PM, Sylvain Lebresne <span dir=3D"ltr">&lt;<a h=
ref=3D"mailto:sylvain@datastax.com">sylvain@datastax.com</a>&gt;</span> wro=
te:<br><blockquote class=3D"gmail_quote" style=3D"margin:0 0 0 .8ex;border-=
left:1px #ccc solid;padding-left:1ex;">
<div class=3D"im">On Mon, Dec 5, 2011 at 8:19 AM, Boris Yen &lt;<a href=3D"=
mailto:yulinyen@gmail.com">yulinyen@gmail.com</a>&gt; wrote:<br>
&gt; Hi,<br>
&gt;<br>
&gt; I am using 0.8.7.<br>
&gt;<br>
&gt; I was trying to use DynamicComposite column. After I intentional added=
 a<br>
&gt; column (string:string:uuid) into a record which has previous columns<b=
r>
&gt; inserted with comparator (string:uuid:uuid). I got an exception on the=
<br>
&gt; cassandra side and an timeout exception on the client (I was actually<=
br>
&gt; expecting an validation exception on the client).<br>
<br>
</div>By design, the DynamicComposite comparator cannot return a validation=
<br>
exception in that case. Or rather, since there is nothing specifying how<br=
>
a column should look like, the only way we could ever hope to return a<br>
validation exception would be to scan the entire row on insertion (which<br=
>
is out of question).<br>
<div class=3D"im"><br>
&gt; After that when I re-started cassandra, cassandra kept=A0saying that<b=
r>
&gt; &quot;comparator mismatch while comparing two dynamicCompositeType col=
umn...&quot; and<br>
&gt; cassandra never really got started.<br>
&gt;<br>
&gt; Is this a bug?<br>
<br>
</div>No. As said above, this is something we cannot do anything about.<br>
The DynamicCompositeType comparator is by nature a very dangerous<br>
one. One column inserted in the wrong row can corrupt that row (with<br>
recovery involving losing data).<br>
In other words, don&#39;t use DynamicCompositeType. Don&#39;t use it unless=
<br>
you are absolutely sure you won&#39;t ever insert the wrong column in the<b=
r>
wrong row *and* you absolutely need the DynamicCompositeType (by<br>
opposition to the faster, cheaper and safe CompositeType). And if you<br>
do, you have to assume the consequence if you do mess up.<br>
<span class=3D"HOEnZb"><font color=3D"#888888"><br>
--<br>
Sylvain<br>
<br>
--<br>
Sylvain<br>
</font></span></blockquote></div><br>

--0015174c18384a0dc004b35535a3--