Re: DB Appliance: the attack of the clones

> # Namely in case of web supporting databases, 99.9% uptime requirement
> is almost always mandatory. And that means RAC.....
>
> Some people engineer ways to "go offline" periodically and temporarily
> disconnect while keeping web applications working. Queue stuff up
> somehow/somewhere and process it asynch after the database is back.
>
> I would argue or at least assert that working to keep some kind of
> reasonable periodic maintenance windows ( aka allowable downtime ) is
> a desirable place to get to.

The thing that needs to be established is the difference - or rather: similarities - between HA and DR.

I'd rather have a DR site that I can push into main operation, while patching the main site. For that, I need Dataguard/GG. Not RAC. And that gives me DR as well - something RAC doesn't.

> Remember a key concept of Moans Nogood is that typically when you
> drink the koolaid of RAC for 99.9 you usually end up with 99.5 or
> 98.2. Complications, troubleshooting skills, bugs, maintenance
> windows ... it never ends in the RAC world.

Now,now! Too much reality here! Enough frivolity, please!

> Maybe 10 years from now the rolling upgrades and patching stuff will
> be there. It is getting there now ... but getting there is a long
> ways from actually being there now and deployable and maintainable by
> everyone.

I was at Roland Slee's introduction of 9i in Australia, when it was presented as "mostly for RAC". I asked him, very clearly: "does RAC let me upgrade each node concurrently and without having to shutdown?". Note that this was the first release of 9i! Without batting an eyelid, he said: "yes, of course! It works perfectly! That is the whole purpose of RAC!"

Yup! Sure...
Any wonder why I don't believe ANYTHING they say until I see proof in real installed sites?

> Just yanking your chain mostly!

Nothing wrong with that. It might actually make someone at Oracle open their eyes? Received on Mon Sep 26 2011 - 06:04:04 CDT