Hi,
Could someone please explain to me in an idiots guide kind of way!!! How
Active Active clustering works when the underlying Windows 2003 cluster
is Active\Passive using mscs with a central SAN.
I dont understand how you can run SQL on both nodes when only one node
can see the actual SAN resources at anyone time.

The scenario I have is I currently have a Windows 2003 Active/Passive
cluster with SQL 2000 running on 2 nodes with a SAN. We want to utilise
the other node to be able to run a another high end database that we are
developing to run along side the first one. However if one of the nodes
fail the both Databases should be able to run on one node until the
other is running again. For scalability purposes we may at some point
wish to add a third node if this is possible.

If someone can explain this to me and how to implement it from both
Windows point of view and SQL I would be grateful.

many thanks

A/A is not a SQL term, but it is a Windows Server 2003 Clustering term. That
part can be very confusing. I have no idea why SQL does not use the term,
but I will explain it anyways

In SQL you have instances of SQL. You can have 16 per cluster. 1 instance
will own the resources at a time. It's part of the shared nothing model. You
can run two or more instances on a node at a time. The other node does not
have to have any instances. Or you can run both nodes with 1 instance each,
giving one master each, and as many User DB's as needed each.

Cheers,

Rod

MVP - Windows Server - Clustering
http://www.nw-america.com - Clustering
http://msmvps.com/clustering - Blog

"Jody Stoll" <jo (AT) mastercare (DOT) co.uk> wrote

Actually it is pretty simple. You install a complete second instance of SQL
with all new disk devices, IP addresses, and SQL Instance Name. Each
database instance is completely independent. The only issue you have is
making sure there is enough resources available, especially physical memory,
when you 'stack' the instances on the same host.

--
Geoff N. Hiten
Microsoft SQL Server MVP
Senior Database Administrator
Careerbuilder.com

I support the Professional Association for SQL Server
www.sqlpass.org

"Jody Stoll" <jo (AT) mastercare (DOT) co.uk> wrote

Geoff,
thanks for the swift reply, is there any chance you can detail the steps
that I have to go through in order to achieve this. Both nodes have
sufficient resources to run both databases on one node when that happens.
However , i'm still vague on the actual steps needed in order to achive this
as well as the steps taken with regards to the windows side of things as i
still dont understand how the 'passive' windows node can access the Shared
drives when it is 'active' as far as sql is concerned.

cheers






"Geoff N. Hiten" <SRDBA (AT) Careerbuilder (DOT) com> wrote

"Jody Stoll" <jo (AT) mastercare (DOT) co.uk> wrote

Either node can see resources on the SAN, but a resource can only be seen by
one node at a time.

In other words, if you create 6 LUNS on your SAN, you can make 3 available
to Node 1, 3 to Node2, or 4 to Node 1, 2 to Node 2, or even 6 to Node1 and
none to Node 2.

So, it's really a matter of how you slice up your SAN.

For example, in my current cluster I have the 3 disks in a Resource group
dedicated to SQL server on Node2 and my Quorom disk in a resource group on
Node1.

So, I think to answer your questions... carve more LUNS out of your SAN and
create a new resource group (with the disks, IPs, etc), move that to your
second node and go from there.

"Jody" <jo (AT) mastercare (DOT) co.uk> wrote

They use DIFFERENT drives.

Close, but that won't work in a cluster. The LUNS need to be visible to all
nodes in the cluster. The cluster service arbitrates ownership so only one
node can access each disk resource at a time. Active-Active is a holdover
from the SQL 7 clustering days when clustering was sort of hacked into the
product. The proper term is multi-instance.

Anyway, SAN resources are typically presented as one or more LUNS. Each LUN
maps to a physical disk in Windows and becomes a physical disk resource for
the cluster. Each LUN is managed independantly, unless you group them
together in a cluster resource group. That would be used for keeping
several disks together for use by a SQL instance.

--
Geoff N. Hiten
Microsoft SQL Server MVP
Senior Database Administrator
Careerbuilder.com

I support the Professional Association for SQL Server
www.sqlpass.org

"Greg D. Moore (Strider)" <mooregr_deleteth1s (AT) greenms (DOT) com> wrote

"Geoff N. Hiten" <SRDBA (AT) Careerbuilder (DOT) com> wrote

I wasn't clear.

You're right the LUNS have to be visiable to all nodes, but only one at a
time can "own" it. That's what I meant about "make visible"

Sorry for the confusion.

Hi,
The installation you have in the default instance that
run on the primary node. binaries on C:\ databases on
shared cluster disks.

You have to add a new shared clustered disk to the OS an
then to the cluster, Disk resource.

Then, on secondary Node, Node2, you install a named
instance to the local C:\ (new explicit set of binaries
and executables) and the databases to the new sahred disk.

Now you have to intances, default (node1) and named
(node2) with databases on external disks. Then you
install two Vitual SQL Servers (separate IP addresses)
for each instance. You can add additional named instances
to any other node, if yoy add more disks. With multiple
instances SQL200x also support a local installation on
the local disk of ONE node.

Don't even try this. Maybe in SQL 7 this would work. All this will do with
SQL 2000 is confuse you and break your system.

--
Geoff N. Hiten
Microsoft SQL Server MVP
Senior Database Administrator
Careerbuilder.com

I support the Professional Association for SQL Server
www.sqlpass.org

"Sca" <anonymous (AT) discussions (DOT) microsoft.com> wrote