Brocade FastIron Ethernet Switch Stacking Configuration Guide - Manual (Page 104)

• After a switchover or failover, the Syslog may contain invalid (non-existent) port numbers in
messages such as “Interface portnum state up.” This is because some messages from the old
active controller will remain in the Syslog after a switchover or failover.
• Failover for devices connected to the management port is not supported. For example, if during a
failover, an end station is connected to the stack through the management port of the active
controller, the connection will be shut down. After the failover, the management port on the new
active controller will work.
• The following describes hitless stacking limitations with software-based licensing for BGP:
‐ To enable BGP on a stack unit, you should have an appropriate BGP license installed on
all the stack units.
‐ If the active controller has a BGP license but any other unit in the stack does not have,
you cannot enable BGP on the stack unit.
‐ If the active controller is not running BGP, a stack unit is operational regardless of
whether the active controller or stack units have a BGP license or not.
‐ If the active controller is running BGP, and a unit without a BGP license joins the stack,
the unit is put into a non-operational state. But, If a user copies the BGP license to a
non-operational unit, it must take effect immediately and becomes operational. Or, if the
user disable BGP, active controller will again put all the non-operational units in
operational mode.
What happens during a hitless stacking switchover or failover
This section describes the internal events that enable a controlled or forced switchover to take place in
a hitless manner, as well as the events that occur during the switchover.
Real-time synchronization among all units in a stack
Hitless stacking requires that the active controller, standby controller, and stack members are fully
synchronized at any given point in time. This is accomplished by baseline and dynamic
synchronization of all units in a stack.
When a stack is first booted and becomes operational, baseline synchronization occurs across all of
the units in the stack. The active controller copies the current state of its CPU to all units of the stack,
including the standby controller. The information received from the active controller is programmed
locally in the hardware on all units. The information includes:
• Start-up and runtime configuration (CLI) - These files are copied to the standby controller only.
• Layer 2 protocols - Layer 2 protocols such as STP, RSTP, MRP, and VSRP run concurrently on
both the active and standby controller.
• Hardware Abstraction Layer (HAL) - This includes the prefix-based routing table, next hop
information for outgoing interfaces, and tunnel information.
• Layer 3 IP forwarding information - This includes the routing table, IP cache table, and ARP table,
as well as static and connected routes.
• Layer 3 routing protocols are not copied to any of the units in the stack, but remain in standby
state on the standby controller until a switchover occurs. Peer adjacency will be restored after a
switchover. If BGP4 or OSPF graceful restart are enabled during a switchover, the standby
controller (new active controller) will initiate a graceful restart and a new set of routes will be
relearned. The new set of routes will be the same as the old routes, except in the case of a
network change.
• OSPFv2 and OSPFv3 non-stop routing support - For more information, refer to the FastIron
Ethernet Switch Layer 3 Routing Configuration Guide.
When control protocols are synchronized and protocol synchronization timers have expired, the
standby controller will be in hot-standby mode, meaning the standby controller will be ready to take
What happens during a hitless stacking switchover or failover
102 FastIron Ethernet Switch Stacking Configuration Guide
53-1003090-02