MSA 1500cs (Active/Active) - SQL Clustering - Drives on 2nd Node not working.

[Mark Ervin]

Hi,

I'm having trouble trying to configure an SQL Cluster using the following;

MSA 1500cs (Active/Active)
2 x MSA 30
1 x MSA 20
2 x BL460c
Windows 2008 R2
SQL 2008 R2

The first server 01 is up and running and the 4 drives are presented fine. The cluster is on line and all databases are working. Ive built 02 in the same way and am using MPIO in windows to mount the drives. Everything appears OK apart from the drive status is an exclamation mark with:

Offline (The disk is offline because of policy set by an administrator)

If I leave the drives as they are I can restart the server fine and carry on as normal. If I manually bring the drives online the server will not boot agaian until I remove the access in the configuration utility. Another thing to note is the drives say reserved on the 01 sql box when looking at the properties.

Is it possible to have 2 nodes accessing the same luns on the MSA 1500cs?
Could it be there is some kind of lock on the MSA?
Is this a clustering issue or an HP issue?

Pretty stuck on this so any help would be appreciated.

Thanks,

Mark

[Johan Guldmyr]

Hello, what file system do you have on the shared disks?

[Mark Ervin]

Hi,

The drives are NTFS. They are configured using windows MPIO. I have tried using the HP DSM in the MPIO manager to no avail.

To add, once the drives are first bought online manually after being presented they appear to come online but no drive letters are given or availabale to assign. It is after restarting then that the server will not boot.

Thanks,

Mark

[Johan Guldmyr]

Hi, you should not access the same NTFS disk from two machines. If it works it will most likely cause corruption.
Check out CSV - cluster shared volumes. There may be other options as well.

http://en.wikipedia.org/wiki/Cluster_Shared_Volumes

[Mark Ervin]

Thanks for the link. It looks as though it may work as a solution for me. My only question is that as SQL will be running in active / passive so only one server will be accessing the disks at once. It is the initial mounting of the drives to setup the cluster I need to get around.

Are you saying it is just not possible or there are better ways of doing it?

Thanks again for your advice. I really appreciate it.

[Mark Ervin]

Also these are stand alone servers and not in a virtual environment.

Thanks,

Mark

[Johan Guldmyr · Accepted solution]

"My only question is that as SQL will be running in active / passive so only one server will be accessing the disks at once. It is the initial mounting of the drives to setup the cluster I need to get around."

OK the application is active/standby, but the accessing/mounting of the disks are not - right? I mean if you have an application with data on a shared disks, even in an active/standby mode wouldn't the standby server need to query the shared disks sometimes? Maybe for other reasons if you want to do a failover it would speed things up if the server already had the disk available. If it doesn't, then the application would need to be told by the other machine when to tell the OS to mount/unmount the disk. Sounds complicated.

With NTFS it may be technically possible for multiple physical servers to access the same NTFS disk.

However, I've seen setups with windows 2003 where they did this and the filesystems did get corrupted.

I'd assume it's the same in 2008. 

In 2008 I believe there is this program/test you can run to see if your cluster is properly set up/if your clustering works.

I don't remember the name of it though.

[Mark Ervin]

Thanks again for your help. True, the Quorum will need to be seen by both so disks will need to be mounted at all times. I think I will just replicate the data and set up the Cluster so it isnt using shared disks.

Thanks for your answer.

Mark

[Johan Guldmyr]

Oh yeah that's one way to do it I suppose.

Glad to help.

// Johan