Irf Multi-Active Detection (Mad) Mechanism - H3C S5820 Configuration Manual

IRF uses a strict configuration file synchronization mechanism to ensure that the members of an IRF
virtual device can work as a single device and that when the master fails, the other devices can operate
normally.
When a slave starts up, it automatically searches for the master, synchronizes the master's
configuration file, and executes the configuration file. If all members start up simultaneously, the
slaves synchronize the master's initial configuration file and execute it.
When the IRF virtual device operates normally, all your configurations are recorded into the current
configuration file of the master, and are synchronized to each member device. When you save the
current configuration file of the master as the startup configuration file by using the save command,
all slaves execute the same saving operation to make the startup configuration file of all members
consistent.
Through the real-time synchronization, all members keep the same configuration file. Even if the
master fails, all the other devices can execute various functions based on the same configuration file.
2) Configuration file application
The configuration file can be divided into two parts: global configuration and port configuration. When a
slave applies these two kinds of configurations of the master, it handles them as follows:
Global configuration: All slaves execute the current global configuration on the master. In other
words, all members apply the same global configuration.
Port configuration: When a slave applies the port configuration of the master, it cares about the
configuration related to its own port. For example, the slave with member ID 3 only cares about the
configuration related to the GigabitEthernet 3/0/x port on the master. If there is a configuration
related to its own port, it applies the configuration; if not, no matter what configuration has been
made to the port before the slave joins the IRF, the slave functions by using the default
configuration.
IRF topology maintenance
Direct neighbors of an IRF virtual device periodically exchange hello packets (the period is 200 ms).
Without receiving any hello packet from a direct neighbor for 100 periods, a member considers that the
hello packets timed out, and the expired device is isolated and the topology database is updated.
When an IRF port of a member is down, the member immediately broadcasts the information to all the
other members. If the IRF port of the master is down, an election is triggered.

IRF Multi-Active Detection (MAD) Mechanism

A link disconnection causes an IRF virtual device to divide into two or more IRF virtual devices with the
same global configuration, which may cause a network failure. Therefore, the multi-active detection
(MAD) mechanism is introduced to detect whether devices on the two sides of the detected link belong
to the same IRF virtual device and then judge whether the IRF virtual device is already partitioned into
multiple IRF virtual devices. The MAD mechanism provides the following functions:
Detection: Enabled for an IRF virtual device, the MAD mechanism detects the network for multiple
active IRF virtual devices with the same global configuration. This is done with the Link
Aggregation Control Protocol (LACP) or the Bidirectional Forwarding Detection (BFD) protocol.
Collision handling: When an IRF virtual device is partitioned, if multiple active IRF virtual devices
are detected, the MAD mechanism keeps only the one with the lowest master ID to operate
normally (keeping the active state). The state of all the other IRF virtual devices are set to recovery
1-10