Drives in a cluster environment

Hi,

I have a SAN and configuring a cluster on SQL 2005. I initially created a Quorum drive when setting up the cluster and now added 4 more drives to the physical node but when I try to install SQL that drive cannot be located.

Do we need to create all the drives when installing the cluster or what is the way to add the drives later on.

Thanks

Anup

I remember that adding a drive for sql requires the drive to be added to the resource group, I assume you have to create a resource group prior to install?|||

It's one thing to add them locally - did you add them to the cluster via Cluster Admin as well? There's a procedure for adding drives after the cluster is already configured. Are they seen in CluAdmin? If not, you did not add them properly.

You also didn't mention MS DTC. That needs its own drive now as well in its own group with an IP, name, and DTC resource.

|||

Thanks All.

With some troubleshooting and tips from this forums the cluster is up and running with SQL 2005. However I do not understand whi we need a seperate drive for MSDTC cant we share the quorum drive to do this. Can you explain how to size this like perf issues, disks needed etc.

Thanks

|||

For failover clusters, you always need to think in terms of "units of failover". If you put the MSDTC data on the same drive as the quorum, you would need to tie MSDTC to the cluster service. You then create dependancies between MSDTC and the cluster service that shouldn't be there.

In general, you want to think in terms of a service and its associated resources. That bundle needs to be independant from all other bundles on the cluster so that it can move from node to node independant of other bundles (resource groups).

For disk storage, the unit of availability is the physical disk (or LUN in the case of a SAN array). You can't have one partition of a disk mounted to one node and another partition mounted to another node.

So, it's not a matter of capacity or throughput or perf issues, it is an availability issue.

|||

Kevin Farlee wrote:

For failover clusters, you always need to think in terms of "units of failover". If you put the MSDTC data on the same drive as the quorum, you would need to tie MSDTC to the cluster service. You then create dependancies between MSDTC and the cluster service that shouldn't be there.

Just to reinforce - this is not a recommended configuration. It only existed with Windows 2000 because of comclust and that has (thankfully) been taken away in W2K3. Always put MS DTC in a separate group.

In the cases where MS DTC is heavily used (such as BTS), it could potentially affect the availability of the quorum disk (i.e. if it gets filled up with DTC log). You don't want to go there.

|||

Assuming you added the new drives as a disk resource in cluster manager, did you also make those new drives a Dependency of SQL Server? If not, SQL Server will not be able to see them.

|||

Thanks. Now I am getting to understand the concepts. However one last question if I have a active/active cluster then do i need to setup 2 MSDTC groups one in each node?

I appreciate all the help throughout .

|||

No, one DTC per cluster. It is shared with everything else in the cluster.

You will need separate dedicated disks and such for your other SQL instance(s) though.