I'm looking to implement a "not quite" hub and spoke type architecture using AlwaysOn and wondered how the cluster quorum might play into this scenario if a remote site can't connect to the hub where we setup a file share witness or the other nodes in the underlying cluster. If the remote location was the primary for the AG and had a local secondary and also a remote secondary (Corp), would the AG automatically failover to the remote secondary even though there may not be an issue inside the remote location?
For example: The hub we will call corporate and the remote locations we will call stores, all on the same AD domain. I put all twelve stores (two servers each, a primary and a local secondary) and the corporate server (will be a central secondary to each of the remote primary servers) into the cluster together (not using Failover Clustering, just the backbone of AlwaysOn) and setup a file share witness. Now let's say that store 1 looses WAN connectivity back to corporate for a few hours. Since neither of the servers (primary or local secondary) can connect to any other store (node) and the file share witness can't be reached, will the Availability Group automatically failover if I have it set for Automatic Failover and Database Level Health Detection? Or will it understand that it has gotten separated from the cluster and simply do nothing?
I realize this is a bit of a complex use case when it comes to AlwaysOn, but thought I'd see if anyone had knowledge of how this scenario might work.
