Specifications
Distribution Layer April 2014
68
Step 3: Configure the system processes to use the loopback interface address for optimal resiliency:
snmp-server trap-source Loopback 0
ip ssh source-interface Loopback 0
ip pim register-source Loopback 0
ip tacacs source-interface Loopback 0
ntp source Loopback 0
Procedure 4 Configure IP unicast routing
The single logical distribution layer design, when configured with VSS, uses Stateful Switchover and Nonstop
Forwarding to provide subsecond failover in the event of a supervisor data or control plane failure. This ability
reduces packet loss in switchover to redundant logic and keeps packets flowing when the data plane is still intact
to adjacent nodes. In the stack-based distribution layer approach, a single logical control point still exists and the
master control plane in a stack can fail over to another member in the stack providing near-second or subsecond
resiliency.
When the supervisor or master switch of a distribution platform switches over from the active to the hot-standby
supervisor or switch, it will continue switching IP data traffic flows in hardware. However, the device in the active
role requires time to reestablish control plane two-way peering with IP routing neighbors and avoid the peer
router from tearing down adjacencies due to missed hellos that would cause a reroute and disruption of traffic.
To allow this time for the device taking over the active role to recover, there is a Nonstop Forwarding (NSF)
setting for the routing protocol to wait for the dual supervisor peer switch to recover. The neighboring router is
said to be NSF-aware if it has a newer release of Cisco IOS Software that recognizes an NSF peer. All of the
platforms used in this design are NSF-aware for the routing protocols in use.
The distribution layer switch is configured to enable NSF for the routing protocol in use so that it can signal a
peer when it switches over from a previously active to a hot-standby device, to allow the peering neighbor time
to reestablish the IP routing protocol relationship to that node. No tuning of the default NSF timers is needed in
this network. Nothing has to be configured for an NSF-aware peer router.
Option 1: Conīgure EIGRP unicast routing
Enhanced Interior Gateway Routing Protocol (EIGRP) is the IP unicast routing protocol used in this design
because it is easy to configure, does not require a large amount of planning, has flexible summarization and
filtering, and can scale to large networks. If you use OSPF as an alternative to EIGRP, choose Option 2.
Step 1: Enable EIGRP named mode for the IP address space that the network will be using. If needed for
your network, you can enter multiple network statements. Enable all routed links to be passive by default.
The Loopback 0 IP address is used for the EIGRP router ID to ensure maximum resiliency. Because routing
functionality is bounded at the distribution and not extended into the access layer, every distribution is configured
as a stub network, optimizing performance. The summary keyword allows summary routes to be advertised and
summarization is used whenever possible.
ip routing
!
router eigrp CAMPUS
address-family ipv4 unicast autonomous-system 100
af-interface default
passive-interface
exit-af-interface
network 10.4.0.0 0.1.255.255