R3303-HP HSR6800 Routers IRF Configuration Guide

7
IRF multi-active detection
An IRF link failure causes an IRF fabric to split in two IRF fabrics operating with the same Layer 3
configurations, including the same IP address. To avoid IP address collision and network problems, IRF
uses multi-active detection (MAD) mechanisms to detect the presence of multiple identical IRF fabrics,
handle collisions, and recover from faults.
Multi-active handling procedure
The multi-active handling procedure includes detection, collision handling and failure recovery.
Detection
The MAD implementation of this device detects active IRF fabrics with the same Layer 3 global
configuration by extending the LACP, BFD, or gratuitous ARP protocol.
These MAD mechanisms identify each IRF fabric with a domain ID and an active ID (the member ID of
the master). If multiple active IDs are detected in a domain, MAD determines that an IRF collision or split
has occurred.
You can use at least one of these mechanisms in an IRF fabric, depending on your network topology. For
a comparison of these MAD mechanisms, see "Configuring MAD."
Collision handling
When multiple identical active IRF fabrics are detected, MAD compares the member IDs of their masters.
If the master in one IRF fabric has the lowest member ID among all the masters, the members in the fabric
continue to operate in Detect state and forward traffic. MAD sets all the other IRF fabrics in Recovery
(disabled) state and shuts down all their physical ports but the console ports, physical IRF ports, and any
ports you have specified with the mad exclude interface command. The Detect-state IRF fabric is active.
The Recovery-state IRF fabric is inactive. Only members in the Detect-state fabric can continue to forward
traffic.
Failure recovery
To merge two split IRF fabrics, first repair the failed IRF link and remove the IRF link failure.
If the IRF fabric in Recovery state fails before the failure is recovered, repair the failed IRF fabric and the
failed IRF link.
If the active IRF fabric fails before the failure is recovered, enable the inactive IRF fabric to take over the
active IRF fabric and protect services from being affected. After that, recover the MAD failure.
LACP MAD
LACP MAD requires that every IRF member have a link with an intermediate device, and all these links
form a dynamic link aggregation group, as shown in Figure 6. In additi
on, the intermediate device must
be an HP device that supports extended LACP for MAD.
The IRF member devices send extended LACPDUs with TLVs that convey the domain ID and the active ID
of the IRF fabric. The intermediate device transparently forwards the extended LACPDUs received from
one member device to all the other member devices:
• If the domain IDs and the active IDs in the extended LACPDUs sent by all the member devices are
the same, the IRF fabric is integrated.