Mount Points and SQL

[NiCkNaMe]

Guys,

My PSU has died in my lab so I can't test this out myself at the moment

SQL 2k8r2 cluster consisting of multiple nodes, with multiple instances

If using mount points for the cluster disks, then is a volume mount point root (e.g. the disk that holds the mounted volumes) needed per instance?

I gather the answer is yes, as the disk would be a dependency of the instance and if you wanted to fail that instance over, and you had other instances using that volume mount root they would also fail over as it would be their dependency too.

So in a multi-instance clustered environment, when using mount points you need a volume mount root, per instance.

Hope this makes sense

Cheers,

 

[brainchylde]

You need disks and IPs per instance. So multi-instance requires multiple disks and multiple cluster IPs.

You should think about the instance as the cluster, rather than the physical nodes.

Hope that helps?

 

[oddjob62]

If using mount points for the cluster disks, then is a volume mount point root (e.g. the disk that holds the mounted volumes) needed per instance?

Correct, although obviously it can be a very small volume.

Had a 3rd party complain about my SQL cluster because:

them: "the F drive that holds our database is only 1gig, can you make it bigger as it is definitely what is impacting the performace of our (badly designed) database"

me: "how exactly do you think your 500gig database is fitting on a 1gig volume?"

Was rather amusing

 

[ASE001]

Sometime when creating the cluster some mounted volumes will not be seen as a valid clustered resource automatically. To get round this just add them manually. Make sure you set the correct dependency for each mounted volume to that of the mount volume.

Be careful with some applications that (i.e. Blue Print, Lexmark) auto-create dbases, they can check the root folder for the necessary space. So in-effect it checks the mount volume, which is typically 1Gbyte or less. So the installation fails, there are ways a round this?

Saying that they are extremely flexible, I have typically 12 SQL instances on my two node cluster designs (60 volumes). You can get as many as 20-23 SQL instances in total, but you will need SQL enterprise to go beyond 16.