While flipping through the Microsoft Azure website recently I ran across this page which talks about HA/DR configurations for SQL Server VMs. Seeing this I figured they’d be talking about AlwaysOn Availability Groups, or even SQL Server FCIs and using SIOS’s DataKeeper Cluster Edition.
Thankfully the article does talk about AGs (and database mirroring), but doesn’t mention FCIs. What it does talk about at the bottom is how you need to setup GRS (Globally Redundant Storage) to use that are part of your DR plan. What GRS is, is a blob store configuration that stores a copy of the blob in another data center. If you are counting on this as being part of your DR plan, you have already failed. You shouldn’t ever be relying on GRS for DR for your SQL Server VMs, or really any VMs. If you need to be able to fail over to your DR system is anything that resembles a quick timeframe spinning up new VMs and attaching existing storage to them isn’t the way to do that.
More traditional DR plans like the other ones which are talked about in the article (as well as SQL Server Clustered Instances) are going to be a much more useful DR configuration.
The post DR Planning in Azure Requires More Than Replicating VMs to Another Data Center appeared first on SQL Server with Mr. Denny.
I like to think of the GRS approach as a DR plan for your DATA, but not for your systems. Meaning that in the case of a very large regional failure (multiple data centres), one could still get to a copy of their data. But using them for failing over of an entire system during a simple local failure wouldn’t work. No HA there at all.