0

We are looking to set up a failover cluster for our SQL 2008R2 database. To accomplish this, we have 2 servers, both equipped with Fusion I/O cards that will both have identical copies of our production databases on them.

From everything I've read, we're supposed to set the servers up so they are both pulling the databases from a shared file storage of some sort. We intended to have both servers simply run the databases from their respective Fusion I/O cards rather than from a file share.

Clustering seems to be the correct choice for a high availability SQL setup, given the "LooksAlive" and "isAlive" checks it makes, but if the shared storage is a requirement, we may need to consider an alternative solution.

Given that we only really need the HeartBeat network functionality and don't need any shared storage, can we still use a Failover Cluster for SQL?

RockiesMagicNumber
  • 123
  • 1
  • 5
  • you need shared storage but you 'may' be able to turn the fusion cards into shared storage with something like starwinds virtual SAN. It would not be a supported solution by any stretch and I wouldn't suggest doing it. – Drifter104 Oct 14 '15 at 21:11
  • Yeah, that does sound like a recipe for disaster. In an ideal failover situation, the "isAlive" check on Primary fails, and the cluster fails over to Secondary and Secondary already has the mdf/ldf files mounted thanks to its own FusionIO card. – RockiesMagicNumber Oct 14 '15 at 21:22

2 Answers2

0

How will you keep the data in sync across the two FusionIO cards? By it's nature, clustering in 2008R2 requires shared storage. If this is a supported configuration by FusionIO, they should provide the steps to configure correctly.

Jim G.
  • 2,607
  • 1
  • 18
  • 19
  • We run nightly jobs that sync the data - these databases aren't accepting much user input, and for whatever reason if we have an outage mid-day, it's acceptable loss. We'd likely implement better incremental backups so in the event our Primary server fails, we could restore the incrementals to the Secondary server after the failover is complete. I'm reading up more on the FusionIO failover documentation still, thank you for that suggestion. – RockiesMagicNumber Oct 14 '15 at 21:17
0

Look into getting SQL 2012 (or higher). This enables you to have Availability Groups. This stores the database on local storage and replicates changes (sync or async, you choose) to the other instance. It looks like this is a better solution to your problem.

Availability Groups also use MS Clustering to throw the IP/Listerner from one node to another but don't require shared storage. You can still use a quorum disk, but a file share witness will work also.

Bart De Vos
  • 17,761
  • 6
  • 62
  • 81