MAC address withdrawal is a feature that is used to inform other nodes that certain FDB MAC address entries should be immediately unlearned, rather than waiting for them to age out. Traffic destined to these unlearned MAC addresses is then flooded until the MAC addresses are learned again. MAC address withdrawal is enabled by default when the MPLS Feature Pack License has been applied to the switch, but can be disabled. When this feature is disabled, traffic destined to MAC addresses previously learned over a now unusable pseudowire (PW) is not flooded until the FDB entries eventually age out. However, it can take the VPLS network longer to adjust than when MAC address withdrawal is enabled. This section describes how this feature operates when it is enabled.
After certain network recovery events, MAC addresses should be unlearned. For example, if the MTU's last usable transport LSP for the primary spoke pseudowire goes down, the MTU decides to activate and switch to the secondary pseudowire and send the primary and secondary core PE nodes an LDP notification message containing the appropriate PW status codes. The core PE node of the primary pseudowire flushes its FDB of any MAC addresses learned from the MTU, but does not send any MAC address-withdraw messages. It is the responsibility of the MTU to initiate the sending of MAC address-withdraw messages. If MAC address withdrawal is enabled, the MTU sends a MAC address-withdraw message to the core PE node of the secondary pseudowire. This node, in turn, sends its own MAC address-withdraw message to all the other core PE nodes in the VPLS causing each of these other core PE nodes to flush their FDB of any matching learned MAC addresses specified in the address-withdraw message. By withdrawing MAC addresses immediately, the other core PE nodes are forced to flood traffic destined to the MAC addresses specified in the address-withdraw message. If an alternate VPLS path exists, the new path can be quickly learned without having to wait for the FDB MAC entry to age out.
When a node needs to withdraw a MAC address, it can signal the MAC withdrawal for a VPLS using an address-withdraw message in one of two ways: the MAC address is explicitly specified in a MAC TLV; or an empty MAC TLV is sent indicating that all MAC addresses not learned from that node should be unlearned. Because this information may be propagated to multiple VPLS nodes, a control plane processing trade-off exists. To reduce the processing load, ExtremeXOS sends an empty MAC TLV. Additionally, ExtremeXOS supports the processing of multiple withdraw messages per VPLS, since other vendors may choose not to send an empty MAC TLV.