ONTAP Discussions

clustered MS SQL physical servers (with MSC) and MetroCluster

ple
4,955 Views

Hello,

Clustered MS SQL physical servers (2 nodes with MSC)  is it supported in MetroCluster Architecture (1 SQL node in site 1 and other in Site 2) ?

Where should be placed the MSCS Quorum? how to avoid the split brain ?

Philippe

5 REPLIES 5

radek_kubka
4,955 Views

Hi,

This is mainly about a Microsoft stretched cluster & depends on the OS version. E.g. for 2008:

http://technet.microsoft.com/en-gb/library/cc731739.aspx

Your best bet will probably be on Node and File Share Majority, with the witness file share placed in a 3rd location.

For the MetroCluster itself you would normally trigger full site fail-over manually - there were some scripted automated solutions (using 3rd site with a witness), but I think support for it was always an a case by case basis.

Regards,
Radek

ple
4,955 Views

Hi Radek,

Thanks for your answer but i Think the answer is : NetApp OnCommand Site Recovery Plug-in 2.0 for Microsoft.

Your opinion ?

Philippe

radek_kubka
4,955 Views

Hmm, I've seen it briefly a while ago, but then forgot about it.

I'm still not sure about fully automated site fail-over, but the documentation says it is possible:

https://library.netapp.com/ecmdocs/ECMP1120726/html/frameset.html

It's a bit of a mystery for me how it can be protected against a split-brain scenario without a 3rd site.

DAMIEN_ALBIEZ
4,955 Views

Phillipe.... did you get further with this?

I have a test environment similar to what you describe:

  • Stretched cluster with a file share witness in 3rd site.  I have had 2008R2 thru 2012 R2 for Os and Sql up to 2014.  Not much difference between them.
  • On command Site recovery Manager plug-in for the cluster (odd name given the Vmware product)
  • Snap mirror replicating storage between sites.

I have had a good experience with it generally although I managed to break the test Dbs a couple of times forcing fail over.  I did mange to repair them without data loss in each case but it did not inspire me much.  I have had better experience with emc and Hp equivalents in the past.  I haven't yet migrated any live Sql instances to the cluster for this reason.  Finding anyone at Netapp who understands what a physical windows server is, has been challenging!

My current issue and reason for the post is performance related.  My current Vm production Sql server performs better than the physical box (using some queries I use in our monitoring system) both with Luns from the same Sas aggregate.

I have read loads about getting the partition alignment right for netapp storage on a Windows OS but haven't yet been happy with the performance differential between physical and virtual. The blades are Ucs B series , all with late firmware and latest drivers.  Have you had similar experience?

Cheers

Damien.

ple
4,955 Views

Hello Damien,

I don't have similar experience, but if you have performance differences between physical server and VM, I think 2 things can explain this :

The ESX cache and memory VS your physical Server memory and cache (I think this point make the difference) and may be, like you say, the partition alignment between the 2 environments . you must do the partition alignment and NetApp SnapDrive can help you for your SQL physical servers..

Regards,

Philippe

Public