I am trying to install SQL Server 2000 on a two-node Windows 2000 cluster.
I have a few questions about the procedure:

1. When I installed the Cluster service, I gave DBCLUSTER as the network
name for the cluster. When I enter that as a virtual server in SQL Server
2000 setup, I have a message saying that the virtual server already exists.
Do I absolutely have to create a new virtual server with another name and IP
address? The cluster is only for SQL Server 2000, no other services.

2. Just to check things up further in the installation procedure, I created
another virtual server, but when I got to the shared drive selection for the
database, I selected the partition I had created for the databases and I got
a warning saying that I should not configure my databases on the same drive
as the quorum drive. How can I get around this? Should I create a RAID
level logical drive specifically for the Quorum drive, so that the resources
will be separated? Right now, I have a resource called "Disk E: F: G:". E:
is the quorum disk, F: is for the databases and G: for the transaction logs.

3. Is there a step-by-step guide to installing SQL Server 2000 in a
clustered environment? I checked on Microsoft's web site, but only found
one for SQL Server 7.0.

Thanks!

Question 1:

You must a separate virtual server name for each virtual
server you create. This virtual server name must not be
used already on your network. This is the name that will
be resolved to an IP address that will be used by the SQL
instance. The IP address must also be unique.

Question 2:

You can ignore that message and proceed to use the quorum
disk for the SQL2000 instance. However, that is strongly
discouraged because the quorum is critically important to
the operation of the cluster and should be used for any
other purposes.

Note that the cluster service uses the disk signatures of
the drives on the shared bus and stores disk signatures in
the cluster registry. Drive letters mean nothing to the
cluster service and are not used to track disks. In your
case, your E, F, and G drives are on a single disk to the
cluster service. You should add another disk to house your
database files. This disk can be a real disk, a hardware
RAID, or a LUN presented from a SAN. Note that drive
letters are at the filesystem level, whereas the cluster
service talks SCSI, even when it's Fibre Channel storage.

Question 3:

Alan Hirt wrote an excellent white paper on SQL Server
2000 Failover Clustering that contains step by step
information.
http://www.microsoft.com/technet/pro...sql/deploy/con
feat/failclus.asp

The SQL2000 BOL is another important source of info.

As soon as you understand the basics, SQL2000 clustering
is quite straightforward and simple to install/manage. The
SQL7 days were a nightmare, though, for those who were
brave enough to venture into SQL7 HA.

Linchi

Answers Inline

"Eric Caron" <ecaron (AT) nospam (DOT) quebecaffaires.com> wrote

Yes. You need a new virtual server separate from the quorum group for SQL.
You can force it to use the quorum group, but it is against best practices
and will compromise the stability and availability of your cluster.

drive should hold cluster quorum files only. You may need to add more
physical shared drives or repartition your shared drive array at the
physical level.

The SQL Server 2000 Resource Kit has an excellent section on setting up a
high-availability failover cluster for SQL server. That is where I
recommend you start.

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

About question 2, I use a SCSI disk enclosure for the shared bus. If I
define a logical drive at the hardware level, will Cluster service pick it
up as a different drive? For budget reasons, we cannot have a second RAID
array for this (I think the drives in RAID 5 are safe enough for the
operation of the cluster anyway).

Thanks for the reference, it looks very detailed. As for clustering being
easy, I agree with you. Just having SQL Server setup detect that the server
is running in Cluster mode is a nice way to go. In fact, the biggest
problem is that it's the first time I'm doing this, so I don't want to do it
the wrong way. Since the system is not live yet, I'm trying to figure out
the best way to build the cluster so as to not run into problems later when
it's gonna be in production. I will follow the advice in the document you
pointed me to.

Thanks!

"Linchi Shea" <linchi_shea (AT) NOSPAMml (DOT) com> wrote

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

Ok, I will rethink my namespace to give a more generic name for the cluster
itself and a specific name for the sql virtual server itself.

Ok, this is what I wanted to know before tearing down everything and
starting over fresh.

Thank you for your help!