3

We have two 3 node clusters in separate data centers, let's call them ClusterP and ClusterS.

ClusterP is our primary and ClusterS is our secondary for disaster recovery. Each cluster has multiple failover cluster instances on it but they are completely segregated from each other.

We'd like to introduce AlwaysOn Availability Groups from ClusterP to ClusterS. I've read multiple guides to this and there is one question I can't really answer. We do NOT have the ability to have cross-datacenter shared storage (no SAN mirroring across datacenters). With this setup, is it possible to setup AlwaysOn Availability Groups?

I understand if we had a standalone server in the secondary datacenter, we could setup an AG with an FCI by adding that server as a node in WFCS, removing it as an owner from any of the FCIs on ClusterP, and setting it up as an AG member. This method does not use shared storage based on my research.

My confusion comes from having two-three node clusters and how they can work together with AGs. It seems to me I need to disband the cluster on ClusterS, add the nodes to ClusterP, add the FCIs needed then setup the nodes so that the physical nodes in location of ClusterS can only host those FCIs and the nodes in location of ClusterP can only host those FCIs. Will this work?

I'd appreciate any tips.

3

We'd like to introduce AlwaysOn Availability Groups from ClusterP to ClusterS. I've read multiple guides to this and there is one question I can't really answer. We do NOT have the ability to have cross-datacenter shared storage (no SAN mirroring across datacenters). With this setup, is it possible to setup AlwaysOn Availability Groups?

No, that is not possible. Outside of a migration, an availability group cannot straddle multiple failover clusters (WSFC).

If you want to leverage an AG, you'll have to ensure that all participating replicas reside within the same WSFC cluster. As for SQL Server failover cluster instances participating in the AG, that is not a problem. A FCI can be an AG replica.

You can achieve a geographically dispersed cluster with a few different strategies. If you need to have a witness vote (which, if you have three node clusters then that wouldn't help and you'd be best with a node majority quorum model) you could utilize a file share witness.

If what you're looking for is your availability group to span two data centers, then yes that is possible. The only requirement for this situation would be a single WSFC cluster (which is possible to span multiple data centers).

  • Thanks, I'm trying to follow your answer as best as I can. So my last statement seems to be correct. I need to destroy the ClusterS and add those nodes to ClusterP. My confusion comes from not having the ability to do SAN failover across the datacenters. If I set it up so only the nodes in the location of ClusterP can host nodes there and only the nodes in the location of ClusterS can host nodes there am I avoiding SAN failover? Then at that point I can setup the AG, right? – Brad May 30 '14 at 17:23
  • You won't need any shared storage for the nodes with an odd number of nodes or a file share witness. And the nodes participating in the availability group won't have any shared storage, as we're talking about non-shared storage between replicas. So you could design your cluster and AG so that isn't a problem. Let me know if that still doesn't make sense. – Thomas Stringer May 30 '14 at 19:32
  • @ Thomas Stringer I am planning to add a 3rd node from a different site to an existing 2 node AAG setup which has Node and File share majority. Can you please let me know if I need to move the File share to the remote data center or Disable the voting rights of the 3rd node as there will be even number of votes? The 3rd node being planned will be in a different subnet . – user2923332 Feb 23 '17 at 21:26

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.