TMS zl Management and Configuration Guide ST.1.1.100226
8-5
High Availability
Overview
3. The TMSVLAN settings that were configured on the cluster participant
before becoming a cluster member are permanently erased.
4. When the cluster master fails, the cluster participant becomes the cluster
master without significant interruption.
5. When the former cluster master comes back online, it uploads the startup-
config of the current cluster master and becomes the cluster participant.
Because of this behavior, you must take care when configuring and rebooting
cluster members to avoid losing all of your settings.
For example, suppose that you already have one TMS zl Module installed and
configured in its host switch, and you want to install a second, unconfigured
module for HA and you take the following actions:
1. Configure the second module for HA and reboot it
2. Configure the original module for HA and reboot it.
Then the unconfigured module comes online first as the master. When the
original, configured module comes online, it detects the cluster master and
reboots using the empty startup-config. Now all of the original module’s
settings are lost.
Linking Inter-Chassis Clusters
An inter-chassis cluster consists of cluster members that are physically
located in different switch chassis. When connecting the two chassis, follow
these guidelines:
■ There should be at least one Layer 2 link and preferably redundant links,
between the two switch chassis.
■ If you use redundant links, enable the switches’ STP (Spanning-Tree
Protocol) or RSTP (Rapid Spanning-Tree Protocol) functionality to pre-
vent a loop.
■ The switch ports that connect the host switches should be tagged mem-
bers of the HA VLAN.
■ All TMS VLANs must be configured on both host switches.
■ It is highly recommended that other network devices not be placed
between the switch chassis that contain the cluster members to reduce
the risk of link failure.
If there is only one link between the host switches of an inter-chassis cluster,
and that link fails, the cluster participant cannot receive the “heartbeat”
messages from the master. The participant therefore assumes that the master