I am running a 3active/1passive node cluster on W2K3 and SQL 2k. The passive
node has the ability to have any of the active nodes to failover to it. In
addition to it being the failover node it also has 3 instances that match the
active nodes.

Since all the drive letters must be shared between the 6 instances I have
quickly run out of drive letters. Each node currently has 2 logical drives
accounting for 12 of the letters, add in local drives and cd drive,A-F and
Z:. I have 5 letters left.

I would like to add a seperate drive on each instance for the system dbs but
do not have enough letters left. Is there a way around this restriction?

Hi

No. Volumes without drive letters are not supported on Clusters.

Having a match of a DB to a drive letter does not gain you much, unless you
know each drive letter is a separate LUN and separate drives on the SAN.

We operate our SANS with 3 drivers per instance, a Data, a backup drive and
a log drive. All RAID-10. Each one is at SAN (EMC) level on different
drives.

Regards
--------------------------------
Mike Epprecht, Microsoft SQL Server MVP
Zurich, Switzerland

IM: mike (AT) epprecht (DOT) net

MVP Program: http://www.microsoft.com/mvp

Blog: http://www.msmvps.com/epprecht/

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

It sounds like you are mapping SQL 7.0 behavior to a SQL 2000 4-node
cluster. There are a lot of differences. First, each instance no longer
has a designated home node. You can limit the nodes an instance is allowed
to run on, but you are far better off allowing all nodes to be hosts for all
instances. You can move instances around on nodes as necessary. For
example, SQL_Instance_1 normally runs on Node A but can run on any node.
SQL2 and SQL3 normally run on nodes B and C respectively but also can run on
any other node. The second choice of all instances is node D, the normally
empty node. Once you have a failover (or just an instance move for
maintenance) you will need to rearrange the failover order to keep from
accidentally stacking the instances on the same node.

Each LUN should also map to a single drive letter cluster-wide. For
example, SQL_Instance_1 should use data drive R: and Log Drive S: on all
nodes. With three instances, you will have plenty of drive letters even with
the cluster overhead.

Geoff N. Hiten
Microsoft SQL Server MVP.


"Mike Epprecht (SQL MVP)" <mike (AT) epprecht (DOT) net> wrote

Just to clarify I have 6 instances on a 4 node cluster. 3 of those instances
reside on one server most of the time while the other 3 are spread across the
other servers?

If drive letter A-F and Z are used for local drives on each of the nodes
that only leave 18 drives for the instances, Q is for quorum. At this point
I can only add 1 more drive letter to each instance or 6 drives total. Since
currently each instance has two drives. After I use these 6 drive letters
will there be any way to add additional drives to the instances.

I am wanting to keep my data spread out as much as possible to avoid Disk
I/O bottle necks.

"Thom" wrote:

You are limited to only named drives a-z with SQL Server 2000. There is no
supported way around it. Can you consolidate some of your local drives to
make those letters availble for shared drives?

Cindy Gross, MCDBA, MCSE
http://cindygross.tripod.com
This posting is provided "AS IS" with no warranties, and confers no rights.