RE: DataGuard

I am reading this list every day and I must say that the general mood of most interveners is that RAC, at least low end RAC, have a bad record in term of HA. Many expressed the opinion that a single instance has a better HA and this is due to the weakness of complexity of RAC. I will add that low end RAC, typically 2 nodes have less skilled people working on it. I assume that 'Yahoo' which runs a 32 nodes RAC has a bunch of highly skilled RACmen.

But we can only express opinion, since there is no statistics, official or unofficial on RAC performances versus single instances.

If my boss rush in the room and ask me to swear that RAC is better in term of HA, I would be annoyed. On one side RAC is supposed to be more HA but on the other side its complexity gives him bad reputation.

In my current portfolio of DB to manage, I have a newly born RAC, not stressed at all (sleeping in progress... please wait) and around 100 single instances, up to now I have no problem with the RAC once it is setup. On Unix, last year, among the 100 db, I had to shutdown 2 of them due to excessive ORA-4031 and we had 2 box crash - I give free all windows box reboot due to ORA-4030, but we are speaking of real server, isn't it Nial? That's all for the unexpected events to mark for 2006 for 100 db. I have already posted on this list a DB 8.0.6, single not OPS, which is up since 2002 on Solaris box and still running fine.

So If I had to responds on the spot to my boss request, based on past tracks, I would say, not worth the money and the troubles. One could argue that the 4 failures we encountered would have been avoided with a RAC. True, but unknown in this count is how much RAC crash due to clusterware, cache fusion specifics, interconnect specific?

In fact, I am implementing RAC but I am not convinced that service will be better. Boss read Oracle propaganda, boss pay me, RAC is good in my contractor CV, so let's do it. Will it be better.... no idea.

Bernard Polarski
Oracle DBA  

-----Original Message-----
From: Mladen Gogala [mailto:mgogala_at_verizon.net] Sent: dinsdag 16 januari 2007 14:07
To: cjpengel.dbalert_at_xs4all.nl
Cc: rgoulet_at_kanbay.com; oracle-l_at_freelists.org Subject: Re: DataGuard

On 01/16/2007 02:42:58 AM, Carel-Jan Engel wrote:
> Mladen,
>
> If a network cannot keep up with the redo of one instance, how could
RAC
> improve that?
>

Carel-Jan, I thought of RAC as the first level of failure protection. If RAC fails,
there's still a standby database. That was what I found interesting in the idea
of the original poster.

-- 
Mladen Gogala
http://www.mladen-gogala.com

--
http://www.freelists.org/webpage/oracle-l




--
http://www.freelists.org/webpage/oracle-l