Brocade FastIron Ethernet Switch Stacking Configuration Guide - Manual (Page 61)

When a stack breaks into partitions, the partition with the active controller remains operational. If a
partition contains the standby controller, this partition will become operational because the standby
controller will assume the active role and will reload the partition units. If the hitless-failover enable
command is configured, the partition units are not reloaded. A partition without an active or standby
controller will not function. To reconfigure these units to act in standalone mode, you must first enter a
stack unconfigure me command on each unit.
To reverse the partitioning, reconnect all of the units into the original stack topology using the stacking
ports. This is the same as merging stacks. If the original active controller again has the highest priority,
it will regain its role. If two partition active controllers have the same priority, the active controller with
the most stack members will win the election. This process helps minimize traffic interruption.
Ring topology stacks do not partition in the event of a member failure. Operation is interrupted briefly
while the stack recalculates a new path. Ring topologies are more stable than linear topologies because
they provide redundant pathways in case of accidental failure.
Merging traditional stacks
Traditional stacks may be merged, but the total number of stack units must not exceed eight. For
example, you could combine two stacks with four units each into a single stack of eight units.
You can merge stacks by connecting them together using the stacking ports. Before doing this, make
sure that none of the stacking ports have been reconfigured as data ports (for example, ports on an end
unit in a linear stack topology). You cannot use secure-setup to merge stacks because secure-setup
does not work across stack boundaries.
When stacks are merged, an election is held among the active controllers. The winner retains its
configuration and the IDs of all of its original stack members. The remaining stack units lose their
configuration and are reset. If the IDs of the losing stack units conflict with the IDs of the winning units
they may change, and the IDs will no longer be sequential. You can use secure-setup to renumber the
members in the newly merged stack. The following examples show how stack merging works:
• If a stack partitions into multiple stacks because of a connection failure, and you fix the connection,
the stack partitions will merge back into the original stack with no change to stack IDs, because in
this case all stack IDs are distinct.
• In a linear stack topology, the end units of the stack will have only one stacking port configured.
Before you can merge two linear stacks, you must reconfigure the end units so that both ports are
stacking ports.
MIB support for the traditional stack
All statistics about packets received and sent, RMON, jumbo frames, runts, giants, and other instances
are gathered through the stack interfaces and are accessible through SNMP. The functionality for
traditional stack is the same as that for a standard 10 Gbps interface. Information includes types of
modules, including optics modules.
NOTE
A type counter has been added to count the number of packets greater than 1518 bytes (jumbo
frames).
Persistent MAC address for the traditional stack
The MAC address for the entire traditional stack is determined by the MAC address of the active
controller. When an active controller is removed from the stack, and a new active controller is elected,
by default the MAC address of the new active controller becomes the MAC address for the traditional
Merging traditional stacks
FastIron Ethernet Switch Stacking Configuration Guide 59
53-1003090-02