Info on active/active model for SQL 2005 cluster

[micha123]

Hi Guys,
I was looking for info about the active/active model for SQL Server 2005 on Windows 2003 cluster.
All the documentation I found in the last two days of research did not give me any answers.
I just need a diagram or a paragraph explaining how it works. Is it like the active/active model in SQL 2000, meaning two different instances (one on each node), clustered on the other node?
What is the difference between the active/active model in 2000 and 2005?
Please don't send me standard links explaining the whole cluster architecture since I have been there already...
%-)

"A long life is when each day is full and every hour wide!"
(A Jewish Mentor)
M.

 

[Catadmin]

I'm new to clustering myself, and I didn't set up the hardware (we've got separate teams), but here's how we have it set up in our environments. Please forgive vagueness. As I said, I'm new to this.

First, don't know of any differences between 2000 and 2005. SQLBill (I think - If I got the wrong person I apologize in advance) did some 2000 clustering work, so might be able to answer that question.

Our Production environment is supposedly Active/Active. We have one instance on one node which is somehow (not sure how) propogated to the other node. Our shares are set up in cluster administrator so when one node fails, its traffic is transparently forwarded to the other node. Sounds like a active/passive, but we're doing load balancing via a foundry (which is a souped up switched hub/router kind of device), so it's not really.

Our Dev/Test/QC environment is active/active with Test & QC on one node and Dev on the other. We have separate paging files on our D: drive (not C: where the OS files are) for both machines and 3 instances of SQL Server, each named Dev, Test or QC. Each instance points to its own set of drives for backups and data/log files. When one node fails, the db(s) in question along with the instance related drives are failed over to the other node. So if the Dev node fails, it flips over to the Test/QC node and vice versa. Again, all our shares are set up in cluster administrator so they fail over properly too. Shares not in Cluster Administrator tend to disappear when the nodes fail over.

This is the extent of my knowledge on clustering, unfortunately. Does any of this help you out?



Catadmin - MCDBA, MCSA
"No, no. Yes. No, I tried that. Yes, both ways. No, I don't know. No again. Are there any more questions?"
-- Xena, "Been There, Done That"

 

[micha123]

Hi Catadmin,

Thank you very much for the detailed answer!
Is it 2000 or 2005 you are talking about?

I have done a lot of SQL 2000 clustering (so I am able to answer questions about that...), but none in 2005.
Did someone do that? Is it the same thing as in 2000, just using the memory dynamically with AWE?

Thanks again!

"A long life is when each day is full and every hour wide!"
(A Jewish Mentor)
M.

 

[Catadmin]

I'm talking 2005 for our setup. We didn't use clustering on SQL 2000 at all.

AFAIK, Clustering is similiar in both 2000 and 2005 for installation. The one thing I know is different is we had trouble with our install Admin group being in a different domain than us and trying to install a SQL Cluster under our domain while having everything set up inside their domain. Apparently Microsoft deliberately designed Clustering so that you have to install the servers inside the domain they will be used instead of cross-domain-something-or-other for security.

I know what I just said doesn't make a whole lot of sense. But it really happened to us and we had to have the install Admins rebuild the box from scratch, and do it from our domain network connections instead of theirs, before it would work.



Catadmin - MCDBA, MCSA
"No, no. Yes. No, I tried that. Yes, both ways. No, I don't know. No again. Are there any more questions?"
-- Xena, "Been There, Done That"

 

[micha123]

You are right about that. I found this in a white paper:

"Unlike SQL Server 2000, in SQL Server 2005, all service accounts that will administer SQL Server, SQL Server Agent, Full-Text Search, or Analysis Services must be added to a global domain group which is then added to the local Administrators group."

It's the Active Directory they are trying to push through... ;-)

"A long life is when each day is full and every hour wide!"
(A Jewish Mentor)
M.

 

[mrdenny]

Active/Active clusting in the MSCS world means one instance on each node. There is no cross node setup for MSCS.

I'm not sure where Microsoft was going with the whole Group thing. Since you have to run under a domain account anyway.

I beleive that everything has to be in the same domain so that if there is a domain to domain failure (or a domain is removed from the tree for somereason) the cluster can't fail.

Denny
MCSA (2003) / MCDBA (SQL 2000) / MCTS (SQL 2005) / MCITP Database Administrator (SQL 2005)

--Anything is possible. All it takes is a little research. (Me)

http://www.mrdenny.com