Re: RAC clarifications
Date: Thu, 17 Dec 2009 21:23:11 +0000
2009/12/17 Guillermo Alan Bort <cicciuxdba_at_gmail.com>:
> Someone once pointed in this list, that having no scheduled downtime is
> unrealistic... and it would end up with unscheduled downtime. So I suggest
> you reviste that part of the question ^_^
Completely agree with you there.
> As for patches that allow rolling upgrades... well. there aren't that many,
> and mostly no CPU allows it (requires startup upgrade, so...)
I'd dispute that on the CPU's I think I've doing a rolling apply of every CPU this year. To be fair to Oracle, I think they genuinely have made a bit of an effort in reducing downtime with this.
> I'm given to understand that there are some patches that can be applied
> online in 11g... but think of it this way: Applying a patch online means
> changin a binary or something in the dictionary... that could lead to a
> whole lot of trouble and require consistent copies of libraries and
> binaries... this is something very hard to accomplish, if at all possible.
> On the other hand, there is no 0 downtime solution with Oracle. There is a
> 'very little downtime' solution, which is RAC + DG... which has a lot of
> issues still. But still, in the event of the ENTIRE RAC going down, there
> would be a downtime, minimal, but downtime at last.
> As mission-critical systems go, not counting custom systems, Oracle RAC + DG
> is the best money can buy... the question would be... how much are you
> willing to spend...
> A six node RAC with a TAF enabled application would have virtually no
> unscheduled downtime, and minimal scheduled downtime (make sure you use
> separate homes) even a three or four node version of that would work...