Wireless Access

In our current wireless set-up we have created a lot of Layer-2 connected high-availability controller clusters. Where each separate controller is connected to a switch with layer 2 and layer 3 functionality. These switches are in the same site-location and are layer-2 connected and so the layer-2 connection between the controllers in the cluster is established. This works fine. 

All controllers are so called MD's managed by Mobility Master.

But now we would like to make a Layer-2 high-availability cluster with Aruba controllers that are in different site-locations. And the switches which these controllers are connected to don't have a Layer-2 connection between them. So our normal way of forming a Layer-2 cluster between controllers isn't possible in this case.

What we can do, by using a dark fiber that's available between the two site-locations, is connect both the controllers directly to each other using this physical link. 

Question: is it possible to connect the controllers in this way uysing the dark-fiber link and establishing a Layer-2 controller cluster in that way?

By the way: Our Mobility Master and controllers are running on Aruba OS 8.5.0.3 currently.

Possible, yes, but you would introduce a different type of failure scenario.  If two controllers are connected directly, so that their heartbeats succeed, but their other uplinks to the network are not available or intermittently unavailable, you will create a black hole for your client traffic.

Clustering was intended to involve a group of controllers at the same physical site with high speed connectivity between them so they can act as one.  They must also have access to the same connectivity, and low latency both with the outside world and with the access points they are serving.  As you move controllers further and further away from each other, the likelihood that a link, a connection or a piece of your infrastructure will be interrupted and force cluster disconnects and heartbeat misses.  Latency will increase and would decrease performance within a cluster silently.  On top of that these issues could be random and can be increasingly difficult to troubleshoot between datacenters.

Long story short, stick with a cluster at the same physical location and you will minimize issues.  If you want to provide site redundancy, I would have a backup-lms to a second cluster at a different site.