Always On and Failover Cluster Installation

Question

I'm trying to install and setup AlwaysOn on four nodes.  Two nodes in site A, two in site B.  Each pair of nodes is a failover cluster instance.

So I have my first pair of nodes online and running on an FCI in site A.  I've added the two other nodes to the Windows Cluster, so all 4 are now on the same cluster.  The problem I have is that when I try to do a new FCI install for site B, using the default instance name, I get an error about a duplicate resource. 

Is this a known limitation?  Can I not have a default instance in site A and a separate default instance in site B?  Is there something else I'm missing here?

I'm also attaching this article for reference, which discusses the prereqs for this.

http://msdn.microsoft.com/en-us/library/edbab896-42bb-4d17-8d75-e92ca11f7abb#FciArLimitations

Thanks!

Answer 1

I should add, I also read the following from here http://msdn.microsoft.com/en-us/library/ff929171(v=sql.110).aspx.

You can set up a second layer of failover at the server-instance level by implementing SQL Server failover clustering together with the WSFC cluster. An availability replica can be hosted by either a standalone instance of SQL Server or an FCI instance. Only one FCI partner can host a replica for a given availability group. When an availability replica is running on an FCI, the possible owners list for the availability group will contain only the active FCI node.

And I'm trying to decipher correctly.  It sounds like I might not be able to have multiple FCI in the same AG, which would make what I'm trying to do above impossible.  Just not sure I'm reading it correctly.

Answer 2

Hello,

I've added the two other nodes to the Windows Cluster, so all 4 are now on the same cluster. The problem I have is that when I try to do a new FCI install for site B, using the default instance name, I get an error about a duplicate resource.

Since all of the nodes are part of the same cluster, the rules apply for resources. It's not possible to have two default instances (or the same name in general) in the same failover cluster. I would also caution you against this approach as you have an even number of nodes and did not specify that you have an impartial disk or fileshare witness in a 3rd location with redundant links. Should communications between the two sites go down, neither will have quorum (each has 50%).

-Sean

---

Sean Gallardy | Blog | Twitter

Answer 3

Thanks.  Yes, I didn't mention a file share witness, which I would be implementing away from a quorum disk if this model worked, but it sounds like it won't.

I guess this brings up another interesting factor.  When configuring AlwaysOn replication, even between two standalone nodes in different sites, what quorum model is recommended?  It seems to me that node majority would work just fine, unless there's an FCI instance involved, and then it seems like a file share witness would pretty much be required.

Is that the case?

Answer 4

Dan,

I think you might wan tot go back and decide what the requirements are and what combination of techniques and technologies will fulfill this. It's still possible to use the AO (a named instance as these aren't clustered install, these are local installs) but the way your quorum is setup is what controls the failover clusters and AO health checks.

Generally when there are an even number of nodes, Node Majority with some type of tie breaker is used. While this isn't always the case (Some people manually force the cluster up) it's the best case if you would like automatic failover to work as expected. A quick read over different quorum types and writing out a list of possible failures and walking through a thought diagram of how that should work might help as quorum isn't technology based but based on how failures are handled based on the infrastructure implemented.

-Sean

---

Sean Gallardy | Blog | Twitter

Answer 5

Im actually pretty familiar with different quorum configs, just not how they impact SQL AO.  It sounds like I need to treat AO as any other clustered service though.

Thank you.

Answer 6

Back on the SQL side for a moment.  Are you saying that I can have two FCI in a single failover cluster if one of them uses a named instance, and then configure AO between them?  If so, I should be able to get around any issues with this by using an AO listener.

Answer 7

Yes.  You can have a single WSFC with 4 nodes.  Two at each site, with each pair sharing storage.  You can install two different SQL Server Failover Cluster Instances (FCIs), one on each side. They have to have different instance names, but they can use the same drive letter and SQL Directory paths (if you have a high enough Windows patch level).  

This will simplify the AlwaysOn AG operations, as physical database operations like adding files will replicate seamlessly.

See AlwaysOn Architecture Guide: Building a High Availability and Disaster Recovery Solution by Using Failover Cluster Instances and Availability Groups

David

---

David http://blogs.msdn.com/b/dbrowne/