Configuring and Managing FCoE and NPV/NPIV
Domain 3 covers the access edge: FCoE on lossless Ethernet and NPV/NPIV for scaling server connections. Questions often contrast what runs at the edge versus what must remain in the core.
FCoE essentials
- FCoE carries FC frames in Ethernet; FIP performs discovery, FLOGI and login. FCoE is not routable and needs a dedicated VLAN mapped to a VSAN.
- A vFC interface binds to a physical Ethernet interface and is configured like an FC port: mode F, allowed VSANs, and no shutdown.
- The Ethernet path must be lossless: priority flow control, DCBX/ETS, jumbo MTU and consistent VLAN configuration. FCoE traffic uses Ethertype 0x8906; FIP uses 0x8914.
NPV and NPIV
- NPIV lets one physical F port host multiple virtual N ports, each with its own FC ID. It is common for virtualized servers; the core switch performs the fabric login and assigns FC IDs.
- NPV makes an edge switch behave as an N port to the core. The edge does not run a full fabric: no local domain ID, no principal switch role, no local zoning. Server ports are F ports; uplinks are NP ports.
- The upstream switch must support NPIV. NPV traffic maps can pin a server interface to a specific external interface; without a map, the switch selects an uplink.
Worked example
A blade server behind an NPV switch cannot log in. Verify feature npv and npv enable, the server interface is F and in the correct VSAN, the uplink is NP and up, and the VSAN is allowed on both sides. Then check the traffic map and the core switch's NPIV support. If the uplink was configured as E, the edge is trying to be a full fabric switch; that is the fault. The reasoning: NPV only proxies logins when the northbound port is an NP port and the core accepts the virtual N-port login.
Traps
- Running FCoE across a routed boundary or sharing its VLAN with ordinary data.
- Forgetting PFC/DCBX or MTU on one link in the FCoE path.
- Expecting local zoning or a domain ID on an NPV edge switch.
- Configuring the NPV uplink as E instead of NP.