Irf Stack Working Process; Topology Collection; Role Election - H3C LS-3100-52P-OVS-H3 Operation Manual

Not only the physical stack ports of members can be aggregated, but also the physical links between
the stack system and the upper or lower layer devices can be aggregated, and thus the reliability of the
stack system is increased through the link backup.
The stack system comprises multiple member devices: the master runs, manages and maintains the
stack, whereas the slaves process services as well as function as the backups. When the master fails,
the stack system elects a new master immediately to prevent service interruption and implement 1:N
backup.

IRF Stack Working Process

IRF stack management can be divided into three stages: topology collection, role election, and stack
maintenance.

Topology Collection

Each device in a stack exchanges hello packets with the directly connected neighbors to collect
topology of the entire stack. The hello packets carry topology information, including stack port
connection states, member IDs, priorities, and bridge MAC addresses.
Each member records its known topology information locally. At the initiation of the collection, the
members record their own topology information. When a stack port of a member becomes up, the
member sends its known topology information from this port periodically. Upon receiving the topology
information, the directly connected neighbor updates the local topology information.
The collection process lasts for a period of time. When all members have obtained the complete
topology information (known as topology convergence), the stack will enter the next stage: role election.

Role Election

A stack is composed of multiple member devices; each member has a role, which is either master or
slave. The process of defining the role of stack members is role election.
Role election is held when the topology is instable, such as, forming a stack, adding a new member,
stack split, or stack merge. The master is elected according to the following principles one by one, until
the only winner is found out:
The current master wins, even if a new member has a higher priority.
A member with a higher priority wins.
A member with the longest system up-time wins.
A member with the lowest bridge MAC address wins.
In this stage, member ID collision, software version loading and stack merging are also handled, which
are discussed in the later sections.
When a device is booted, it first collects topology information and then participates in the role election.
After that, the stack system can run normally. When the role election is finished, the stack enters the
next stage: stack maintenance.
1-7