Dell SmartFabric OS10 User Guide Release 10.5.4

Single FCF per vFabric

This feature presents all available operational Fibre Channel uplinks in a fabric as a single logical unit. The uplinks are presented as one logical Fibre Channel Forwarder (FCF) to the end points connected to the same fabric.

Whenever the first FC uplink becomes operationally up and completes the initial login to upstream switch successfully, the logical FCF becomes active and is projected to the end devices.

The logical FCF must be projected only if there is at least one operationally up FC uplink; this FC uplink must have successfully completed the initial login with upstream switch at the time of timer expiry. This behavior achieves better load balancing during boot-up and bulk configuration.

Even though all the uplinks are projected as one FCF, when a request for session establishment is received, the system finds the optimally loaded FC uplink. The load balancing algorithm makes use of the link's session count and the link speed as factors for session re-balancing. Session count of the uplink is the count of both Fabric login sessions (FLOGI) and Fabric discovery sessions (FDISC) in that uplink. Link speed will be the speed of the FC uplinks. End devices do not have control over the link chosen for session establishment. This behavior ensures better load balancing across the available uplinks. After the session is established, the FCoE/FC data traffic is re-directed to the appropriate port to which the login request was associated.

Now the logical FCF takes care of the FIP functionality in the VLAN configured for the fabric. With this implementation, all control frames originating from the logical FCF use a system generated MAC address instead of the port's MAC address. This system generated MAC address of logical FCF is same for all the fabrics configured in the gateway switch; because, every FCF is uniquely identified by the end device using VLAN-MAC address pair and the VLAN used is unique for every fabric. As a result, the same MAC address is used for all the fabrics.

The control frames such as FCF Discovery Advertisement and Login Request or Response from logical FCF(s) in a gateway switch have the gateway switch's world wide name(WWN) instead of upstream switch's WWN as Fabric name inside Fabric descriptor. As a result, while forwarding the control traffic from upstream, it is the responsibility of the gateway switch to modify the fabric descriptor in those frames with the logical FCF's fabric name.

When an uplink interface in the gateway switch becomes operationally down, all the sessions associated with that uplink are terminated by sending a session termination request(CVL/FLOGO) to the corresponding end devices. When the end devices requests for re-establishment of those sessions, those devices are allocated to the next optimal (least loaded) link available in the fabric.

The manual load re-balance done using a management interface command has no impact and it continues to do load re-balancing across the upstream ports(instead of FCFs) available in the gateway switch. You must mention the fabric id while triggering and can monitor the load re-balancing in the gateway switch using the corresponding management interface display commands such as show npg uplink-interface.

Restrictions and Limitations

Connecting uplinks of the same fabric in a Gateway switch to two different SAN networks is an invalid configuration. If you connect the same fabric in a gateway to two different SAN networks, the following scenarios may occur:

• There is no conflict in FC address assignment between the fabrics - This scenario leads to reduced visibility. End devices can only talk to the other end devices connected to the same fabric.
• There is conflict in FC address assignment. The following scenarios may further occur:
  • If the conflict occurs for the FC uplink interface's initial login, the corresponding FC session is closed. Initial login is retried till an unique address within NPG fabric context is assigned with a longer retry time out period(10 seconds). Untill login succeeds, this interface will not be a part of the logical FCF.
  • If the conflict occurs for a forwarded login request (FLOGI/FDISC), the older session with the same FC-ID survives and the newer session is teared down or rejected. This behavior is notified to the user through logs visible to the customer. The reason for the error is set and the duplicate FC Id counter variable is incremented.