A UCS Fabric Interconnects pair can be configured in a high availability cluster. In the cluster configuration, FIs create a primary secondary affiliation to control data traffic flow whereas the data plane forwards simultaneously on both Fabric Interconnects. This means cluster control traffic flow is controlled only by the primary node whereas network data can flow through both FIs under normal conditions. Fabric Interconnects are connected through dedicated Gigabit Ethernet UTP copper ports, which do not participate in the data plane. These interfaces are marked as L1 and L2. L1 of one Fabric Interconnect should be connected to the L1 of the other Fabric Interconnect and similarly L2 should be connected to L2 of the peer. These links carry cluster heartbeat traffic.

High availability configuration requires three IP addresses. Both Fabric Interconnects require an IP address each, and the third IP address is the cluster IP for the management that floats between the peers A and B depending on which peer is primary.

Under normal conditions, both Fabric Interconnect peers must have the same hardware platform and software version. For example, a UCS 6332 cannot be peered with 6332-16UP. However, this requirement may temporarily be allowed during an FI upgrade. Peers are upgraded one by one, which means that for a brief period, they can have different software and hardware.

Once cluster physical cabling is complete, both FIs go through the initial setup process. During the initial setup of the first Fabric Interconnect, it must be enabled for cluster configuration and it is configured as primary and assigned a cluster management IP. When the second FI is configured, it automatically detects the cluster peer and is configured as secondary.