Skip to content

SQL2012 Clustering – the drive letter is gone!

January 26, 2012

For the past few days, I’ve been digging into some of the new features of the upcoming SQL2012 release (currently at RC0), especially around the new HA functionality called Availability Groups and around clustering setup/management differences in general.

One change in particular, that actually has nothing to do with either of these, caught my attention and that I just had to try out. SQL2012 now supports locating database files on a network file share. This new option is documented in 2012 BOL, and Brent Ozar (twitter | blog) wrote a great followup post on the subject earlier this month as well. Actually, to be fair I can’t truthfully call this a “new option”, as previous versions of SQL Server were able to host databases on file shares with the use of a special trace flag. However, this is the first version that supports it out of the box, start to finish.

For a clustered system, there follows an interesting implication. By placing all of the databases onto a network share – including the system databases such as master, msdb, tempdb, etc – SQL Server no longer requires a drive letter dependency! Which, in turn means there is no longer a limit on the number of instances a cluster can run – or at least no limit due to the 26 drive letter restriction. Whether you’d want to run 27 instances of SQL Server on your cluster is a whole other ball of wax that I’m not even going to touch, but at least now you can do it. Just like you can mount a JATO unit on a 1967 Impala. Just don’t light the darn thing.

But I digress.

So, let’s take this one step further. A cluster does not need a shared drive for quorum if you’re using Node or File Share quorum. In other words, SQL no longer needs shared drives, and Windows doesn’t need a shared drive. Well, to me that’s a cluster without SAN!! Wow, hello 2012. Would I run a production system this way? Not a chance. Development or QA? Sure, why not? All a highly-available development system needs now is another Windows box.

So, the HA/DR picture becomes even more blurry this year. If you have a copy of SQL2012 and two Windows boxes, you now get the fun of deciding among logshipping or mirroring (nothing new there), the new Availability Groups option (mirroring on steroids – too bad they really couldn’t have called it that), and now also clustering without SAN. Oh, and good old transactional replication. Nice. I’m sure somebody much smarter than I can summarize a blog post on the ins and outs of each of those technologies.

A final note before I wrap up. If you do use a file share for hosting your system database, it appears that it needs to be on a Windows2008 server. I could not get SQL to work against a Windows2003 file share, and have filed a Connect bug on the restriction. I have no idea if a SMB share on Linux would work. Again I’ll leave that for people smarter than myself.

Happy 2012 clustering!!


Comments are closed.

%d bloggers like this: