Greetings,

I've been asked to try and take our first foray into setting up a SQL 2005
Active/Passive cluster for a SharePoint deployment. I've not done clustering
before and could ue a few pointers and hopefully get a few questions
answered.

We have 2 identical servers each running 2K3 Server Enterprise 64-bit and
SQL 2005 Standard 64-bit. Both servers are running 2 physical disks in a
RAID 1 for OS and programs, and 3 physical disks in a RAID 5 for data.

I have been reading through truckloads of documentation in order to do this
but am unsure about a few things. On Server 1, the SharePoint databases have
already been setup by an external vendor and SharePoint has been configured
(the Index and Web servers are on separate servers). SQL on Server 2 has
been patched up to the same version as on Server 1 but there are no
databases on it as of yet.

I understand that the very first steps I need to follow are to to install
and configure MSCS and configure a clustered MS DTC. Beyond that it's all
reading through documentation.

Questions:

1) Will I need additional disks? (Quorum/shared resource disks on the nodes,
etc.?
2) Will I be able to cluster these two nodes and leave the existing
databases intact?
3) If yes to #2, do I need to copy the databases to the second server before
I can implement clustering?
4) Anything else I should watch out for?

Cheers and many thanks in advance for any assistance.

Armando

First, you need to share the disks--that your database files will
reside--among the two servers. This can be done either through a common SCSI
bus or a SAN. Not many people use a shared SCSI bus in a cluster these days.

If you make your thee disks into one RAID5 set, you basically have one disk
as far as clustering is concerned. Ideally, you should have at least another
RAID-protected disk for your quorum.

Yes, you can cluster the two servers, install SQL Server, and migrate your
databases to the new clustered SQL instance. Ideally, you should start to
cluster two servers with them in a clean state rather than from their current
state.

You don't need to copy the databases to the second server. You should save
your databases just as when you need to rebuild these servers.

Many people have bad experience with clustering because they don't follow
the documentation and don't devote enough time to gain sufficient experience
in a test environment first. Then, they complain that failover clustering
itself seems to be causing more down time that it is supposed to prevent. I'd
strongly suggest that you allocate sufficient time to planning, learning, and
testing. Don't start with the prod right away.

Linchi

"Armando" wrote:

First, you need to share the disks--that your database files will
reside--among the two servers. This can be done either through a common SCSI
bus or a SAN. Not many people use a shared SCSI bus in a cluster these days.

If you make your thee disks into one RAID5 set, you basically have one disk
as far as clustering is concerned. Ideally, you should have at least another
RAID-protected disk for your quorum.

Yes, you can cluster the two servers, install SQL Server, and migrate your
databases to the new clustered SQL instance. Ideally, you should start to
cluster two servers with them in a clean state rather than from their current
state.

You don't need to copy the databases to the second server. You should save
your databases just as when you need to rebuild these servers.

Many people have bad experience with clustering because they don't follow
the documentation and don't devote enough time to gain sufficient experience
in a test environment first. Then, they complain that failover clustering
itself seems to be causing more down time that it is supposed to prevent. I'd
strongly suggest that you allocate sufficient time to planning, learning, and
testing. Don't start with the prod right away.

Linchi

"Armando" wrote:

First, you need to share the disks--that your database files will
reside--among the two servers. This can be done either through a common SCSI
bus or a SAN. Not many people use a shared SCSI bus in a cluster these days.

If you make your thee disks into one RAID5 set, you basically have one disk
as far as clustering is concerned. Ideally, you should have at least another
RAID-protected disk for your quorum.

Yes, you can cluster the two servers, install SQL Server, and migrate your
databases to the new clustered SQL instance. Ideally, you should start to
cluster two servers with them in a clean state rather than from their current
state.

You don't need to copy the databases to the second server. You should save
your databases just as when you need to rebuild these servers.

Many people have bad experience with clustering because they don't follow
the documentation and don't devote enough time to gain sufficient experience
in a test environment first. Then, they complain that failover clustering
itself seems to be causing more down time that it is supposed to prevent. I'd
strongly suggest that you allocate sufficient time to planning, learning, and
testing. Don't start with the prod right away.

Linchi

"Armando" wrote:

First, you need to share the disks--that your database files will
reside--among the two servers. This can be done either through a common SCSI
bus or a SAN. Not many people use a shared SCSI bus in a cluster these days.

If you make your thee disks into one RAID5 set, you basically have one disk
as far as clustering is concerned. Ideally, you should have at least another
RAID-protected disk for your quorum.

Yes, you can cluster the two servers, install SQL Server, and migrate your
databases to the new clustered SQL instance. Ideally, you should start to
cluster two servers with them in a clean state rather than from their current
state.

You don't need to copy the databases to the second server. You should save
your databases just as when you need to rebuild these servers.

Many people have bad experience with clustering because they don't follow
the documentation and don't devote enough time to gain sufficient experience
in a test environment first. Then, they complain that failover clustering
itself seems to be causing more down time that it is supposed to prevent. I'd
strongly suggest that you allocate sufficient time to planning, learning, and
testing. Don't start with the prod right away.

Linchi

"Armando" wrote:

First, you need to share the disks--that your database files will
reside--among the two servers. This can be done either through a common SCSI
bus or a SAN. Not many people use a shared SCSI bus in a cluster these days.

If you make your thee disks into one RAID5 set, you basically have one disk
as far as clustering is concerned. Ideally, you should have at least another
RAID-protected disk for your quorum.

Yes, you can cluster the two servers, install SQL Server, and migrate your
databases to the new clustered SQL instance. Ideally, you should start to
cluster two servers with them in a clean state rather than from their current
state.

You don't need to copy the databases to the second server. You should save
your databases just as when you need to rebuild these servers.

Many people have bad experience with clustering because they don't follow
the documentation and don't devote enough time to gain sufficient experience
in a test environment first. Then, they complain that failover clustering
itself seems to be causing more down time that it is supposed to prevent. I'd
strongly suggest that you allocate sufficient time to planning, learning, and
testing. Don't start with the prod right away.

Linchi

"Armando" wrote:

First, you need to share the disks--that your database files will
reside--among the two servers. This can be done either through a common SCSI
bus or a SAN. Not many people use a shared SCSI bus in a cluster these days.

If you make your thee disks into one RAID5 set, you basically have one disk
as far as clustering is concerned. Ideally, you should have at least another
RAID-protected disk for your quorum.

Yes, you can cluster the two servers, install SQL Server, and migrate your
databases to the new clustered SQL instance. Ideally, you should start to
cluster two servers with them in a clean state rather than from their current
state.

You don't need to copy the databases to the second server. You should save
your databases just as when you need to rebuild these servers.

Many people have bad experience with clustering because they don't follow
the documentation and don't devote enough time to gain sufficient experience
in a test environment first. Then, they complain that failover clustering
itself seems to be causing more down time that it is supposed to prevent. I'd
strongly suggest that you allocate sufficient time to planning, learning, and
testing. Don't start with the prod right away.

Linchi

"Armando" wrote:

First, you need to share the disks--that your database files will
reside--among the two servers. This can be done either through a common SCSI
bus or a SAN. Not many people use a shared SCSI bus in a cluster these days.

If you make your thee disks into one RAID5 set, you basically have one disk
as far as clustering is concerned. Ideally, you should have at least another
RAID-protected disk for your quorum.

Yes, you can cluster the two servers, install SQL Server, and migrate your
databases to the new clustered SQL instance. Ideally, you should start to
cluster two servers with them in a clean state rather than from their current
state.

You don't need to copy the databases to the second server. You should save
your databases just as when you need to rebuild these servers.

Many people have bad experience with clustering because they don't follow
the documentation and don't devote enough time to gain sufficient experience
in a test environment first. Then, they complain that failover clustering
itself seems to be causing more down time that it is supposed to prevent. I'd
strongly suggest that you allocate sufficient time to planning, learning, and
testing. Don't start with the prod right away.

Linchi

"Armando" wrote:

First, you need to share the disks--that your database files will
reside--among the two servers. This can be done either through a common SCSI
bus or a SAN. Not many people use a shared SCSI bus in a cluster these days.

If you make your thee disks into one RAID5 set, you basically have one disk
as far as clustering is concerned. Ideally, you should have at least another
RAID-protected disk for your quorum.

Yes, you can cluster the two servers, install SQL Server, and migrate your
databases to the new clustered SQL instance. Ideally, you should start to
cluster two servers with them in a clean state rather than from their current
state.

You don't need to copy the databases to the second server. You should save
your databases just as when you need to rebuild these servers.

Many people have bad experience with clustering because they don't follow
the documentation and don't devote enough time to gain sufficient experience
in a test environment first. Then, they complain that failover clustering
itself seems to be causing more down time that it is supposed to prevent. I'd
strongly suggest that you allocate sufficient time to planning, learning, and
testing. Don't start with the prod right away.

Linchi

"Armando" wrote:

First, you need to share the disks--that your database files will
reside--among the two servers. This can be done either through a common SCSI
bus or a SAN. Not many people use a shared SCSI bus in a cluster these days.

If you make your thee disks into one RAID5 set, you basically have one disk
as far as clustering is concerned. Ideally, you should have at least another
RAID-protected disk for your quorum.

Yes, you can cluster the two servers, install SQL Server, and migrate your
databases to the new clustered SQL instance. Ideally, you should start to
cluster two servers with them in a clean state rather than from their current
state.

You don't need to copy the databases to the second server. You should save
your databases just as when you need to rebuild these servers.

Many people have bad experience with clustering because they don't follow
the documentation and don't devote enough time to gain sufficient experience
in a test environment first. Then, they complain that failover clustering
itself seems to be causing more down time that it is supposed to prevent. I'd
strongly suggest that you allocate sufficient time to planning, learning, and
testing. Don't start with the prod right away.

Linchi

"Armando" wrote: