Brocade FastIron Ethernet Switch Platform and Layer 2 Switching Configuration Guide - Manual (Page 27)

NOTE
Global MAC addresses have priority over dynamic flow-based MAC addresses. To ensure that global
MAC addresses are in sync across all packet processors, flow-based MAC addresses may be
overwritten in one or more packet processors. The MAC addresses will be relearned and
reprogrammed using the flow-based method as needed by incoming traffic flows.
Flow-based learning configuration considerations
When configuring flow-based MAC learning, consider the rules and limitations in this section.
• Flow-based MAC learning is not supported with the following features:
‐ Disabling the automatic learning of MAC addresses (CLI command mac-learn-disable ).
‐ Globally disabling Layer 2 switching (CLI command route-only)
• Configuring flow-based mac address learning and Multi-Chassis Trunking on the same device is not
supported.
• When flow-based MAC learning is enabled, unknown unicast packets are copied to the CPU.
Therefore, flow-based MAC learning should not be enabled if a continuous high rate of unknown
unicast packet flooding is expected, as this will cause high CPU utilization.
• Unknown unicast flooding can occur for a known destination MAC address, if the system fails to
program that destination MAC address because the hardware MAC table or hash bucket is full. This
condition can also lead to high CPU utilization.
• A source MAC address is learned only on the ingress (source) packet processor. The MAC address
is added to other packet processors as needed by their incoming traffic flows. During a brief period
until the destination MAC address is successfully added to the hardware MAC table, unknown
unicast flooding is expected on the VLAN.
• When a flow-based MAC address moves, it is deleted from all of the packet processors, then
relearned on each packet processor individually, as needed by incoming traffic flows.
• The software MAC address table in the CPU uses a hashing algorithm. Because hash collisions can
occur and may consume software resources, the FastIron may not be able to support up to 32K MAC
addresses.
• The system can scale up to 32K MAC addresses, however, each packet processor is limited to a
maximum of 16K MAC addresses. This limit still applies, as this is a hardware limitation.
• On SXL with authentication protocols and high traffic, the clear mac and mac flush operations log a
lot of new address messages resulting in high CPU utilization for few minutes.
Configuring flow-based MAC address learning
To configure flow-based MAC address learning, simply enable it globally. If necessary, increase the
capacity of the MAC address table as well.
Enabling flow-based MAC address learning
To enable flow-based MAC address learning, enter the following command at the Global CONFIG level
of the CLI.
device(config)#mac-learning-flow-based
This command enables flow-based MAC address learning. All dynamically-learned MAC addresses are
flushed from the hardware and software MAC tables and are subsequently learned using flow-based
MAC address learning.
Flow-based learning configuration considerations
FastIron Ethernet Switch Platform and Layer 2 Switching Configuration Guide 27
53-1003086-04