In a Unified Computing System (UCS) all the switching functionality is done by UCS 6100 series fabric interconnects. A UCS system can have either one or two UCS 6100 series fabric interconnects. If there is only one fabric interconnect then all the switching is performed by this fabric interconnect, however if there are two fabric interconnect then both perform the job of switching and are connected in a high availability configuration.
Single and dual fabric interconnect topology
In case of single fabric interconnect topology the fabric extender from the blade chassis connects to the single fabric interconnect. Note that two fabric extenders in the same chassis can not connect to the same fabric interconnect, and if they do the fabric interconnect will disable one of the fabric extender links. For connecting the fabric extender to the fabric interconnect you can use one, two or four links from the fabric extender to the fabric interconnect. A topology with three links from the fabric extender to the fabric interconnect is not supported.
In case of dual fabric interconnect topology two fabric extenders from the blade chassis connect to two fabric interconnects running in a cluster. Note that if the fabric interconnects are not in the same cluster the fabric extenders can not associate with both of them and thus this topology will not work. Also remember that one fabric extender in the blade chassis should connect to the first fabric interconnect in the cluster and the other fabric extender should connect to the second fabric interconnect in the cluster. For connecting the fabric extender to the fabric interconnect you can use one, two or four links; three links from the fabric extender to the fabric interconnect is not supported.
In both of the topologies (single fabric interconnect or dual fabric interconnect) if one link of the four links from the fabric extender to the fabric interconnect fails then the UCS system will fall back to two links as three link topology is not supported. However this fall back will happen only when you re-acknowledge the chassis in the UCS manager. If you haven't re-acknowledged the chassis, no pinning is changed, and so the remaining 3 links will still pass traffic. The servers vNIC’s mapped to the failed link will remain down unless failover has been configured. After re-acknowledge is done, the UCS system starts working in two link topology and the server vNIC’s are mapped to one of these two links.