One of my customers contacted me; they had set up High Availability but had a concern. They couldn’t ping all of the IP addresses associated with each of their listeners.
I gave it a go, and sure enough, I could see one of the IP addresses, but not the ‘sleeping’ one, which was on a different subnet. The listener had two IP addresses, one for each subnet.
Then it suddenly made sense that this is ‘by design’. The IP addresses are not online at the same time, since the server can’t host an IP address that’s not in its subnet, so you can only have one ‘up’ at a time.
When the cluster group is owned by one of the nodes, one IP address is up. The other is down, sleeping but ‘healthy’. When the cluster group is owned by the other node, the original ‘sleeping’ IP Address is up, and the original IP address is down.
This was easy enough to show. I induced failover, and then bringing it back up again. It is possible to see the swapover of the IP addresses, and the original ‘awake’ IP address going offline whilst the other comes up.
It is still there; just not accessible since the server can’t host an IP address which is not part of its subnet.