This post covers ENCOR 350-401 · 1.1.b on the CCNP roadmap: High availability: redundancy, FHRP, SSO. In 1.1.a we covered the design tiers; now we look at the “what if a device dies” side of that design.

1. It all starts with a single point of failure

A single point of failure (SPOF) is a part that takes everyone down when it goes. Design means putting a backup next to every SPOF. But a backup is not enough; it has to take over by itself. This post focuses on the last two rows of the table:

What is backed up?How?
LinkDual uplinks, EtherChannel
DeviceTwo switches/routers (a distribution pair)
PowerDual power supplies
Routing pathThe routing protocol finds the alternative path
Default gateway (first hop)FHRP: HSRP, VRRP, GLBP
Control plane (supervisor/RP)SSO + NSF

Routing protocols repair the path between routers on their own. But there is a catch: end devices don’t speak routing.

2. The problem: a PC has only one gateway

You tell a PC “your gateway is 10.0.0.2”. Even with two routers in the distribution pair, the PC knows only that address. If 10.0.0.2 dies, the PC can’t think “what about the other one?”. Either you fix every PC by hand, or you make the gateway address movable between two routers. The second is the job of FHRP (First Hop Redundancy Protocol).

Picture a shared company phone number: 100. Two operators sit behind it. Only one answers the number at a time while the other waits quietly. If the operator falls ill, the other starts answering the same number. The caller never changes the number and never notices. The virtual IP is the “line 100” of the gateway, and the routers are the operators.

3. How HSRP works

HSRP (Hot Standby Router Protocol) is Cisco-proprietary. A summary from Cisco’s HSRP documentation:

  • Roles: One router is active (forwards traffic), one is standby (waits). If the active router misses hellos, the standby takes over and a new standby is chosen.
  • Virtual IP and MAC: End devices use the virtual IP as their gateway, not a physical router address. In HSRP version 1 the virtual MAC is 0000.0C07.ACxx (xx is the group number in hex). Version 2 uses the range 0000.0C9F.F000–FFFF. With N routers there are N+1 IP and MAC addresses.
  • Versions: Version 1 is the default; its hellos go to multicast 224.0.0.2 over UDP port 1985, with group numbers 0–255. Version 2 sends hellos to 224.0.0.102, supports groups 0–4095 and advertises timers in milliseconds.
  • Priority: Default 100. Higher wins. On a tie the higher IP wins.
  • Preempt: Off by default. Without preempt, a router with a higher priority that shows up later does not take over from the current active router.
  • Track: When a tracked object (such as an uplink interface) goes down, the router’s priority drops by 10 by default. The decrement is configurable.

Hello/hold times are commonly known as 3 and 10 seconds (I could not verify that detail in the documentation excerpt above, so treat it as general knowledge). A real failover therefore waits until the active router’s silence exceeds the hold time.

Two traps

1. “I gave it a priority but the other one is still active.” That is what happens with preempt off. You gave R1 110, R2 was already active, and R2 does not let go. Fix: standby 1 preempt on R1.

2. “I added tracking but nobody took over.” Tracking lowers R1’s priority (110 → 90). But for R2 to take the active role, R2 needs preempt too. Without tracking, R1 stays active when its uplink fails and traffic hairpins back through the LAN to R2: it works but takes a pointless detour.

A basic configuration:

interface GigabitEthernet0/0
 ip address 10.0.0.2 255.255.255.0
 standby 1 ip 10.0.0.1
 standby 1 priority 110
 standby 1 preempt
 standby 1 track GigabitEthernet0/1 20

4. VRRP and GLBP

All three show up on the exam. Short differences (the VRRP and GLBP rows are general knowledge):

HSRPVRRPGLBP
StandardCisco-proprietaryOpen standard (IETF)Cisco-proprietary
RolesActive / standbyMaster / backupAVG (coordinator) + AVFs
Load sharingNo (one active per group; balance by hand with different groups/VLANs)No (same idea)Yes: one virtual IP, several virtual MACs

VRRP has preempt on by default while HSRP has it off; this difference is a common exam question. GLBP’s difference: it answers client ARP requests with the virtual MACs of different routers, so both routers carry traffic with a single gateway address. In HSRP the standby router sits idle.

5. Control plane redundancy: SSO and NSF

So far we backed up two separate devices. What about the two supervisors/routing engines inside a single chassis? From Cisco’s documentation:

  • SSO (Stateful Switchover): The standby waits fully initialized and synchronized with the active’s persistent and running configuration. It also keeps protocol and hardware state in sync. When the active fails, the standby takes over and Layer 2 traffic keeps flowing using the existing switching information. Enabled with mode sso under redundancy.
  • NSF (Nonstop Forwarding): Works together with SSO. The goal is to keep forwarding IP packets while the control plane switches over. CEF does this: the standby’s FIB and adjacency tables mirror the active’s, so forwarding continues while routing protocols rebuild the RIB.
  • NSF-capable / NSF-aware: The capable device is the one that restarts and relearns routing information from its neighbors. An aware neighbor holds on to its routes and adjacencies instead of dropping them. That way neighbors see no link flap.
  • Protocols: For BGP, graceful restart capability is negotiated in the OPEN message and NSF-awareness is off by default. EIGRP signals a restart with the RS bit in hellos and awareness is on by default. For OSPF, all neighbors on a segment must be NSF-aware.

Analogy: the captain and the first officer watch the same map all the time (SSO, state sync). When the captain drops, the first officer takes the wheel instantly. The car keeps rolling meanwhile (NSF, forwarding via CEF). All passengers (neighbor routers) are willing to wait: “the driver changed but the car didn’t stop” (NSF-aware).

6. Lab: build HSRP and break it yourself

For this item I made an eight-task HSRP lab that runs in your browser: HSRP gateway redundancy (the lab pages are in Turkish). It has two routers, a switch, a PC and an “internet” router. In order: set up the virtual IP, make R1 active, fail the LAN and watch the PC’s ping survive, get the active role back with preempt, and finally cut the uplink and learn why tracking also needs preempt on R2.

An honest note: the simulator shows HSRP’s election logic (priority, IP tie-break, preempt, track) and its result. Hello/hold timers are not simulated, so the switchover is instant; on real gear there is an outage about as long as the hold time. Only HSRP version 1 behavior exists in the simulator.

7. Which one, when?

SituationChoice
Cisco-only network, simple gateway backupHSRP
Multi-vendor networkVRRP
Want both routers working on one VLANGLBP (or different HSRP active routers per VLAN)
One chassis, two supervisors, nonstop L3SSO + NSF
Routed access layerNo FHRP needed, uplinks are already active (see 1.1.a)

8. Check yourself

  1. Why shouldn’t a PC’s gateway be a physical router IP? (if that router dies the PC knows no other path; a virtual IP moves between two routers)
  2. What are HSRP’s default priority and preempt state? (100, preempt off)
  3. Tracking lowered R1’s priority but R2 didn’t take over. Why? (preempt is off on R2)
  4. Which of HSRP, VRRP and GLBP offers load sharing? (GLBP)
  5. NSF-capable vs NSF-aware? (capable: the device that restarts; aware: the neighbor that waits for it)

Sources

The HSRP and SSO/NSF details are summarized from Cisco documentation. The VRRP/GLBP comparison and the hello/hold values are general knowledge. Next item: 1.2.a, Catalyst SD-WAN control and data plane.