Cluster SQL With Database stored on the local Drive

Question

Cluster SQL With Database stored on the local Drive

Chris Pfeiffer

Grasshopper

Points: 12
More actions
December 13, 2005 at 9:03 am

#364425

Please be kind I am new to the DBA arena. That said I will go on with my question.
I am currently working on developing a cluser for our SQL to run on. Right now we have it on a dual 3.0ghz Xeon that is taking a beating. I am working with IBM to get the Blade Center (so if you have any advise on that as well like if you have had good or bad experinces with them let me know ) When talking to IBM sales reps (they will tell you that a 486 will run SQL) they told me that you could run SQL clustered without a SAN. Well I have not been able to get this done. When I go to install SQL I get an error that says:
The installer has encountered an unexpected error. The error code is 2878. On the dialog ClusterGroupDlg the control DataPath has a possible value : C:\WINDOWS\cluster\Microsoft SQL Server\. This is an invalid or duplicate value.
I am running Windows 2003 Enterprise Edition and trying to install SQL 2005 (i have tried 2000 as well and get a similar error) I feel like it is because I do not have a SAN, the reason behind not using a SAN is this database will never grow to be huge it is just that there A LOT of people using it. Please Help a Beginner OUT!

Thanks in advance
Chris

Viewing 6 posts - 1 through 5 (of 5 total)

You must be logged in to reply to this topic. Login to reply

Ian Scarlett SSC-Insane Points: 23197 More actions · Answer 1

Clustering works by having disk storage that is attached to all of the nodes in a cluster. When a nodes running SQL Server fails, one of the other node takes control of the disk drives, starts up SQL Server and, hopefully, the users don't notice that anything has happened. The C drive is obviously local only to that node, and cannot therefore be used in a clustered environment. You don't need a SAN, but the hardware must be in Microsoft's list of supported hardware.

If you are having performance problems, then clustering is not the answer. Clustering is there to provide hardware resilience, not increased performance.

Chris Pfeiffer Grasshopper Points: 12 More actions · Answer 2

So you are telling me that This cluster will not increase performance. We are trying to add resilience as well as performance but the clustering will not help with performance? Also do you know where I can find information at on how to use a local drive and not a SAN but it sounds like we are going to want a SAN from the way you just described it to me. As you can see I am very new to the area. I am open to any reading that you can suggest!

Thanks

Ian Scarlett SSC-Insane Points: 23197 More actions · Answer 3

Microsoft Press have produced a book 'SQL Server 2000 High Availability' (ISBN 0735619204) that explains clustering in huge detail. Search this site or Google, and you will find plenty of links/articles on clustering.

What I meant by not having to use a SAN was that you can use some other shared disk technologies (e.g. NAS), but you won't be able to use your locally attached C: drive as it can't be shared with other nodes in the cluster.

I think you need to address the performance issue separately. A SAN will probably give you better disk performance, but if I/O isn't your problem, it won't help your overall performance.

Jeff Gray SSChampion Points: 10667 More actions · Answer 4

It is possible that the IBM rep was talking about 'clustering' in the general sense, and not specifically talking about Microsoft Cluster Services.

I believe that Veritas makes a clustering product for windows that does not require a SAN. I think they use some type of replication and failover scheme to accomplish the same or similar effect as MSCS without using the shared-nothing architecture.

I don't really have any experience with Veritas' clustering.

jg

vinny_1973 SSCommitted Points: 1782 More actions · Answer 5

BTW ... dont use a Blade Center for a cluster. We ran into some significant issues when attempting this ... and then got free(ish) hardware from OUR IBM suppliers cos it worked out that they didn't know what the hell they were talking about.

The issue is that they have a single Network backbone, and you then, effectively, setup Virtual NIC's that can access is ... and this cause quite alot of issues in the clustered environment. It also kinda violates the principle of clustering ... as your single point of failure is the Blade Centre itself!?

Take some advice ... 2x Reasonable Spec DL380's, (at least 2 PHYSICAL NIC's each!), nice little NAS for shared storage ... and ignore EVERYTHING the IBM rep tries to say about Blade Centres and Clustering! Blade Centres as Web Farms, Yes ... as a resiliant failover clusters ... absolutely not!!