I would like some advise on my plan to install an active/active sql cluster.
The plan is to build 2 boxes with W2k3 server and SQL2003. I'll have 13
drives on a shared disk storage (two sets of mirrored drives) (two sets of
Raid 5 [4 drives each]) (1 global hot spare). I'll create a 1 Gig logical
drive (Q & P)on each of the mirrored sets (P will be a right-off but will
keep everything looking the same). The Q drive is for the Quorum. So I'll
have C: and D: on each machine, (Q: Emirrored drives F: Raid5 (P: G
mirrored drives H: Raid5. SQL1 will own Q: E: and F:, SQL2 will own P: G:
and H:. E: and G: will be for transaction logs while F: and H: are for the
databases. I'll install instances of sql running on each server and spilt
the databases between the two (we have around fifty). Is this a sound plan,
or have I just wasted my time?

That sounds pretty good. I would seriously look at RAID 10 rather than RAID
5. The difference in write performance can be huge.

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

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

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

I do not understand the Q and P drive assignment.

I have a Q (Quorum) in a seperate cluster resource group (with seperate
ip-address an networkname)
I have a X (MSDTC) in a seperate cluster resource group (with seperate
ip-address an networkname)
The Q and X drive do not come back in the SQL cluster resource group's

You do not mention a seperate cluster resource group for MSDTC. You should
do that.

Gr. Gé

(more info on MSDTC: http://sswug.org/blogging/gbrander/)

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

I am not planning going to use and entire physical disk for the Quorum. I'm
going to make a partition on a mirrored set (that will be a physical disk)
that will be the Q: drive. The P: drive is just my way of keeping
everything looking the same. My plan is to have only 2 cluster groups.
Group 1 be will the Cluster IP, Cluster Name, the Physical Disk (E: Q, the
Physical Disk (F, the MSDTC, and the first instance of SQL. Group 1 will
be owned by server1. Group 2 be will the Physical Disk (G: P, the
Physical Disk (H, and the second instance of SQL. Group 2 will be owned by
server2. Will this work ????

"Gé Brander" wrote:

I too and lost by your wording I think, or maybe cause its Monday here. Are
you saying you will split a LUN into 2 or more partitions and then try to
use different partitions with different nodes? Clustering does not deal with
partitions, only drives. So a node or instance will not be able to share a
drive (and one or more partitions) with another node/instance.

Cheers,

Rod

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

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

Good catch Rodney.

Clustering looks at physical disks (LUNs in SAN-speak). If you partition
the disk, clustering still sees the underlying physical disk.

Also, you don't want your Quorum disk to be part of your SQL Resource group.
That is a Low-Availability approach. Putting MSDCT in with a SQL instance
is an even worse approach. You need one disk for the Quorum, preferably one
disk for MSDTC (although it can be the same as the Quorum disk), and at
least one (preferably two or more) disks per SQL instance. These must be
physical disks or separate LUNs from a SAN device. Anything else will
compromise availability to the point that a cluster won't buy you any higher
availability.

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

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

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

Rodney and Geoff, I admit my terminology was bad. I 'AM' going to have 2
physical disks (LUNs in SAN-speak) per instance of SQL. One for databases
and one for transaction logs. I was trying to see if I could cheat and 'NOT'
use an entire disk for the quorum, but looks like that will not work (or not
work very well) in a multiple instance cluster. So it looks like I'll need
14, probably 15 disks on my shared storage to make this work.

So how about plan B:
Setup one physical disk (probably mirrored) for the quorum and the MSDTC.
Setup one physical disk (mirrored) for the transaction logs for each instance
of SQL. Setup one physical disk (Raid 5 or maybe 10) for the databases for
each instance of SQL. And if I can afford it a global hot spare. Better to
get it right in the planning stage than looking like an idiot trying to get a
bad design to work.

Thanks

"Geoff N. Hiten" wrote:

I like plan B, and not just cause I fully understand it. Question, are your
SQL applications going to use MSDTC? If so, for performance reasons you may
want to have a mirror just for the log and extend the default log size.

Cheers,

Rod

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

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

Rodney, yes I plan on the SQL applications using the MSDTC. Where can I find
info on how to extend the default log size.

Thanks for the heads up.

"Rodney R. Fournier [MVP]" wrote:

First follow http://support.microsoft.com/kb/817064 on each machine BEFORE
you install the cluster service.

Then on each node - open component services - Computers - My Computer -
Properties - MSDTC tab - Capacity = 12 or 16 or anything larger then 4 MB,
close it out. Make both machines the same size log.

Install Microsoft clustering.

Finally follow
http://support.microsoft.com/default...b;en-us;301600

Then Install SQL in the Cluster.

Cheers,

Rod

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

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