Re: RAC for dev env

From: Thomas Roach <>
Date: Tue, 2 Feb 2010 17:30:39 -0500
Message-ID: <>

I will give you a story where this bit me BIGTIME.

We had a Solaris 10 RAC cluster ( for production but the business skimped because they didn't want to pay "all that money" for RAC just for a Dev/QA environment. This was a system that was generating 600+ million dollars in revenue, but they didn't want to spend a few hundred thousand to get RAC. Sounds like the tail wagging the dog, and it was!

So time came around to do security patches and what not. They also wanted CRS Bundle patches. So we patched Dev/QA, no problem. When we went to apply the CRS_BUNDLE patch there was an error with the patchset and it wiped out the inittab. (I didn't know what happened, but eventually uncovered inittab). The upgrade script bombed multiple times. I got Oracle support on the line and we troubleshooted it. 8 hours later we got the systems back online.

The owner of this business segment was pissed. We had a post mortem that was not pretty. She looked straight at me and said, why didn't we see these problems in QA or even Dev? It cost us (a lot of money, more than Oracle RAC licensing for Dev/QA would have cost). I told her that her environments did not match. She said why? I said because no one wanted to spend the money to make the environments match. That shut her up real quick.

Needless to say it was a high priority to make them match and they spared no money to get environments up to speed.

I knew the lesson learned, but it was a lesson for the business. You need to explicitly tell them that unless environments match, there is a risk that not everything can be pretested prior to going into production. Let them decide if it is worth it. If they say no, get it in writing.


On Tue, Feb 2, 2010 at 4:51 PM, Guillermo Alan Bort <>wrote:

> Hi Ram
> I have seen this in several places and it always leads to severe downtime
> in production. non RAC Dev environments could work for functional
> development. For performance and concurrency testing and for staiblity
> testing, you need a RAC environment. Furthermore, I'd suggest you get
> someone with a lot of experience in RAC before actually going to production
> with it. Having a TEST/DEV/QA RAC could help you gain that experience.
> I cannot speak as to licensing, but if you have the DB up, then it doesn't
> matter what you use it for... you still have to have a license, at least
> that is how I understand it.
> One suggestion, if you can't get someone with experience... get a couple of
> virtual machines and toy around with them. You can delete them afterwards,
> or burn them to DVD and hide them under your desk in case of an audit by
> Oracle, and still have a RAC sandbox... you wouldn't be able to perform
> stress tests here, but you can test functionality and the quirks of RAC.
> That being said, perhaps reviewing WHY they want RAC would reveal that they
> actually don't need RAC but a simple DataGuard. I've recently recommended
> moving from a two-node RAC to a single instance on a better server (better
> hardware, 64 bit instead of 32 bit and AIX instead of windows) and setting
> up a DataGuard for fast recovery and minimal data loss
> And just a word of advice... never, EVER, use RAC on Windows. Actually...
> never user windows for a database server...
> RAC is like exadata... it's useful only in very specific situations, other
> than that is just a waste of money and resources.
> hth
> Alan.-
> On Tue, Feb 2, 2010 at 6:30 PM, Ram Raman <> wrote:
>> Hi
>> A decision has been made by upper management to use RAC for production
>> environment and nonRAC for dev/test environments. I would prefer to have a
>> parallel environment which is identical to prod. Is it possible to run a
>> dev environment without having to worry about license. If license might
>> be an issue for dev, can we have one environment where we just install RAC
>> for practice before deploying it in production.
>> Any thoughts or guidances on the issue would be highly appreciated.
>> Thanks,
>> Ram.

Thomas Roach

Received on Tue Feb 02 2010 - 16:30:39 CST

Original text of this message