Users Guide

222 Stacking
If a new switch is added to a stack of switches that are powered and running
and already have an elected stack master, the newly added switch becomes a
stack member rather than the stack master. Use the boot auto-copy-sw
command on the stack master to enable automatic firmware upgrade of newly
added switches. If a firmware mismatch is detected, the newly added switch
does not fully join the stack and holds until it is upgraded to the same
firmware version as the master switch. After firmware synchronization
finishes, the running configuration of the newly added unit is overwritten
with the stack master configuration. Stack port configuration is always stored
on the local unit and may be updated with preconfiguration information from
the stack master when the unit joins the stack.
Information about a stack member and its ports can be preconfigured before
the unit is added to the stack. The preconfiguration takes place on the stack
master. If there is saved configuration information on the stack master for the
newly added unit, the stack master applies the configuration to the new unit;
otherwise, the stack master applies the default configuration to the new unit.
Removing a Switch from the Stack
Prior to removing a member from a stack, check that other members of the
stack will not become isolated from the stack due to the removal. Check the
stack-port error counters to ensure that a stack configured in a ring topology
can establish a communication path around the member to be removed.
The main point to remember when removing a unit from the stack is to
disconnect all the links on the stack member to be removed. Also, be sure to
take the following actions:
• Remove all the STP participating ports and wait to stabilize the STP.
• Remove all the member ports of any Port-Channels (LAGs) so there will
not be any control traffic destined to those ports connected to this member.
• Statically re-route any traffic going through this unit.
When a unit in the stack fails, the stack master removes the failed unit from
the stack. The failed unit reboots with its original running-config. If the stack
is configured in a ring topology, then the stack automatically routes around
the failed unit. If the stack is not configured in a ring topology, then the stack
may split, and the isolated members will reboot and re-elect a new stack