Hello,
After reviewing your setup, the issue appears to stem from a common point of confusion when implementing a three-node switchless storage fabric. The core problem is a mismatch between your physical cabling and the logical IP subnet configuration. In a switchless design, a direct cable can only connect two nodes on the same IP subnet. Your current cabling connects adapters from different subnets, such as hv05 adapter 2 (192.168.101.1) to hv06 adapter 2 (192.168.100.3), which prevents IP communication on that link and causes the cluster network to report a failed state.
To create a valid and resilient switchless storage network for three nodes, you should configure three separate point-to-point networks in a triangular or "ring" topology. Here is a step-by-step guide to reconfigure your setup:
Step 1: Adjust Physical Cabling
Please ensure the nodes are cabled as follows to form a complete ring:
hv04 (Adapter 1) ↔ hv05 (Adapter 1)
hv05 (Adapter 2) ↔ hv06 (Adapter 1)
hv06 (Adapter 2) ↔ hv04 (Adapter 2)
Step 2: Reconfigure IP Addresses
Assign IPs from three distinct subnets, one for each direct link. This allows the cluster to route traffic between all nodes efficiently.
Link 1 (hv04 ↔ hv05): Subnet 192.168.100.0/24
hv04, Adapter 1: 192.168.100.1
hv05, Adapter 1: 192.168.100.2
Link 2 (hv05 ↔ hv06): Subnet 192.168.101.0/24
hv05, Adapter 2: 192.168.101.1
hv06, Adapter 1: 192.168.101.2
Link 3 (hv06 ↔ hv04): Subnet 192.168.102.0/24
hv06, Adapter 2: 192.168.102.1
hv04, Adapter 2: 192.168.102.2
Once you apply these changes, Failover Cluster Manager will automatically detect three healthy cluster-only networks. It's also worth noting the Microsoft Learn pattern you referenced is for a switched deployment (using TOR switches), which is why it shows all nodes on the same two subnets. For a true switchless design, the multi-subnet approach above is the correct pattern.
I believe this will resolve the adapter failures you are seeing.
If this guidance helps you solve the problem, please feel free to hit "accept answer"! :)