Shared Drives in Cluster

Question

Shared Drives in Cluster

Ron Berry

SSC Rookie

Points: 36
More actions
September 7, 2005 at 12:31 pm

#110000

I am trying to determine if the following will cause issues in a multinode cluster.

2 active - 1 passive nodes
all disks are raid 0+1
d: drive on all servers is supporting data for instance 1 on server 1, and instance 1 on server 2
e: drive on all servers is supporting data for instance 2 on server 1, and instance 2 on server 2
f: drive on all servers is supporting data for instance 3 on server 1, and instance 3 on server 3

of coarse there would also be a drive for the quarum and drives for log files, index's, and tempdb

Trying to determine what the best way to build out a multi node cluster with multiple instances, as if each instance gets all its own drive letters you run out real quick with each instance getting at least 4-5 drives

Viewing 4 posts - 1 through 3 (of 3 total)

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

Karl Klingler SSCertifiable Points: 5869 More actions · Answer 1

Hi,

you should be aware that each disk can only be contained in 1 cluster group, which can be owned only by 1 Node. Therefore each disk can be accessed only by 1 node at any given time. To give access to a second Node you would have to switch the cluster group containing the disk to that node, thus denying access to the first node.

On w2k you might be able to use different disks for each instance by assigning them not to a drive letter but to a subdirectory...

regards

karl

Best regards
karl

Rick Carisse SSC Enthusiast Points: 140 More actions · Answer 2

Hi there,

We have 2 nodes with 2 instances of Sql 2000 on Win2k configured active/active as such:

D: on each node is local and stores Sql executables

Cluster resources are:

E = logs for default instance

F = data for default instance

H = logs for named instance

I = data for named instance

By default, node 1 owns the resources (E, F) for the default instance and node 2 owns the resources (H, I) for the named instance. When a problem occurs the resources failover to the other node.

As Karl points out only one node can own the shared resources at any given time.

Be sure to limit the memory useage on each instance to be able to handle one node taking over both instances.

This has ben working well for us.

Hope this helps.

Rick

Ron Berry SSC Rookie Points: 36 More actions · Answer 3

Guess I should have said - the drive letters are reserved for that use, as you both correctly pointed out - only one node can own a drive letter at one time.