We have exadata v-2 quarter rack with 6 cell nodes on linux env and we are adding exadata expansion rack HC with 4 cell nodes.
We will be creating development database also and allocated separate diskgroups for dev env.
I have few doubts here, some are related to exadata and some on generic RAC/database queries:
--> In case if there is any issue on production env and we need to down all the cell nodes then development env will also be down. Hence what is the best recommended setting if we are using same cell servers for prod and dev both env?
--> Since we will have prod and dev database on same compute node so both the databases would be using same CRS services and in case any issue arises and CRS needs to be down then both the DB would be down. Please suggest.
--> We have 2 instances for HA on production env and for dev env we do NOT want HA feature and would like to create stand alone db on the running cluser only. What is the recommended approach in this case?
--> What are the other challenges we might face in such scenario when we keep prod and dev on same cluster and on same db/cell servers?
Thanks for your time and suggestions in advance.
Edited by: 877938 on Jun 4, 2012 1:08 PM
Best practice would be to not mix dev and test environments with production. Have you looked into a ZFS storage appliance for these databases? Or ODAs?
My recommendation would be that you not do what you are planning to do. A little extra hardware will address the issue.
Alternatively, though it will be a larger implication budget-wise, would be to purchase a second Exadata to use as a Data Guard physical standby. Make your primary the failover for dev and test ... use the production standby as their primary.
Actually getting Exadata Expansion Rack is the requirement for Production Database only but since we are getting High Capacity Disks and we have enough disk space (66TB) so we thought to keep Exadata Development environment as well for development purpose only. We will not be setting data guard or making secondary database here.
As Oracle Exadata already provides IORM features to balance CPU and IO load on multiple databases, so why keeping development and production environment on same system is not recommended?
The first problem that comes to mind is that you have nowhere to test patches without affecting your production environment. It's not recommended to have cells within the same cluster running different versions of the storage server software. Also, you won't have anywhere to test clusterware upgrades or patches, since all of the databases share the same cluster. Also, your dev/test databases won't test for any RAC functionality since they'll only be single instance.
Andy said most of what I would have said but the point of Exadata (RAC) is high availability.
The point of Dev and Test is to identify things that cause breakage.
There is a very substantial disconnect between the two. If your testing never causes a failure ... then why bother to test? If developers never try new technologies why bother to develop? If every piece of hardware and software on the planet was bug-free none of us would have a job. Oracle is very very good ... but perfection is still over the horizon. I strongly advise you to not mix these loads. For very little money ... an ODA and a small ZFS will give you what you need to be safe. And if it leaves you with a little spare disk capacity on the Exadata ... don't worry ... it will be utilized sooner than you think.