First Hop Redundancy Protocols

Hot Standby Router Protocol

HSRP operates with an active/standby model (more generally called active/passive). HSRP allows two (or more) routers to cooperate, all willing to act as the default router. However, at any one time, only one router actively supports the end-user traffic. The packets sent by hosts to their default router flow to that one active router. Then the other routers sit there patiently waiting to take over should the active HSRP router have a problem.

This next section of the chapter discusses how HSRP achieves its goal of providing default router redundancy. It progresses briefly through the mechanisms of virtual IP and MAC addresses, failover, load balancing, object tracking, and HSRP versions.

HSRP Virtual IP and MAC Addresses

The HSRP active router implements a virtual IP address (VIP) and matching virtual MAC address. This virtual IP address is part of the HSRP configuration, an additional configuration item compared to the usual ip address interface subcommand. This virtual IP address is in the same subnet as the interface IP address, but it is a different IP address. The router then automatically creates the virtual MAC address. All the cooperating HSRP routers know these virtual addresses, but only the HSRP active router uses these addresses at any one point in time.

Using HSRP protocol messages between the routers, the routers negotiate and settle in either the HSRP active or HSRP standby state. The router with the highest HSRP priority wins and becomes active, with the other router becoming the standby router. If the priorities tie, the router with the highest IP address wins and becomes active. (Also, note that while FHRPs often have only two routers in a group, if using three or more, HSRP uses only one standby router, with the other routers in a listen state, waiting to become the new standby router one day.)

The active router implements the virtual IP and MAC addresses on its interface in addition to its configured interface IP address. Hosts refer to the virtual IP address as their default router address instead of any router’s interface IP address. For instance, in Figure 16-5, R1 and R2 use HSRP. The HSRP virtual IP address is 10.1.1.1, with the virtual MAC address referenced as VMAC1 for simplicity’s sake.

Figure 16.5 All Traffic Goes to .1 (R1, Which Is Active); R2 Is Standby

HSRP Failover

Under normal conditions, with all devices and interfaces working, one HSRP router is the default router, with another standing by. That might happen for months before the standby router needs to take over. However, so that the standby router knows when to act, the two routers continue to send HSRP messages to each other.

HSRP uses HSRP Hello messages to let the other HSRP routers in the same HSRP group know that the active router continues to work. HSRP defines a Hello timer, which dictates how often (in seconds) between successive Hello messages sent by the active router. HSRP also defines a Hold timer, typically more than three times the Hello timer. When the standby router fails to receive a Hello from the active router within the time defined by the hold time, the standby router believes the active router has failed, and begins taking over as the active router.

For example, Figure 16-6 shows the result when R1, the HSRP active router in Figure 16-5, loses power. R2 fails to receive additional HSRP Hellos from router R1 for hold time. At that point, R2, the new active router, starts using the virtual IP and MAC addresses.

Figure 16.6 Packets Sent Through R2 (New Active) After It Takes Over for Failed R1

The figure shows packets flowing from the hosts toward router R2, with no changes on the hosts. The host keeps the same default router setting, referencing the virtual IP address (10.1.1.1). The host’s ARP table does not have to change either, with the ARP entry for the default router listing the virtual MAC.

To direct the Ethernet frames that formerly flowed to router R1 to instead flow to router R2, changes occur on both the routers and the LAN switches. The new active router (R2) must be ready to receive packets (encapsulated inside frames) using the virtual IP and MAC addresses. The LAN switches, hidden in the last few figures, must also change their MAC address tables. Formerly, their MAC tables directed frames destined for VMAC1 to router R1, but now the switches must know to send the frames to the new active router, R2.

To make the switches change their MAC address table entries for VMAC1, R2 sends an Ethernet frame with VMAC1 as the source MAC address. The switches, as normal, learn the source MAC address (VMAC1) but with new ports that point toward R2. The frame is also a LAN broadcast, so all the switches learn a MAC table entry for VMAC1 that leads toward R2. (By the way, this Ethernet frame holds an ARP Reply message, called a gratuitous ARP, because the router sends it without first receiving an ARP Request.)

HSRP Load Balancing

The active/standby model of HSRP means that all hosts send their off-subnet packets through only one router. In other words, the routers do not share the workload; instead, one router forwards all the packets. For instance, back in Figure 16-5, R1 was the active router. All hosts in the subnet sent their packets through R1, and none of them sent their packets through R2.

HSRP does support load balancing by preferring different routers to be the active router in different subnets. Most sites that require a second router for redundancy also use several VLANs and subnets at the site. The two routers will likely connect to all the VLANs, acting as the default router in each subnet. The HSRP configuration settings can result in one router being active in one subnet and another router being active in another subnet, balancing the traffic. Or you can configure multiple instances of HSRP in the same subnet (called multiple HSRP groups), preferring one router to be active in one group and the other router to be selected as active in another.

For instance, Figure 16-7 shows a redesigned LAN with two hosts in VLAN 1 and two in VLAN 2. R1 and R2 connect to the LAN using a VLAN trunking and router-on-a-stick (ROAS) configuration. The two routers define two HSRP groups, one to support each of the two subnets. In this case, R1 wins and becomes active in Subnet 1, while router R2 becomes active in Subnet 2.

Figure 16.7 Load Balancing with HSRP by Using Different Active Routers per Subnet

Note that the design uses both routers and WAN links by having each router act as the HSRP active router in some subnets.

The example surrounding Figure 16-7 raises the question of where to consider using HSRP. You should consider an FHRP on any router or Layer 3 switch interface with an IP address that connects to hosts that rely on a default router setting. If only one router connects to the subnet, you do not need an FHRP, but if two or more connect to the subnet, you benefit from using an FHRP.

HSRP Interface Tracking

Another feature supported by all the FHRPs tracks the operational state of other router features. IOS allows for tracking of interface state, tracking routes in the IP routing table, and other types of objects. When the tracked interface or object fails, HSRP reduces that router’s HSRP priority. With well-chosen priority and tracking settings, you can arrange the HSRP configuration so that when everything works perfectly, one router is active. Later, when something fails related to that router, another router preempts and takes over as the active router.

Figure 16-8 shows one classic failure case that can occur without tracking. In this example, router R1 uses priority 110, with router R2 using 100 (the default), so R1 wins and becomes HSRP active. However, the one WAN link connected to R1 fails. R1 remains the HSRP active router. In this failure case, hosts forward packets to router R1, which has to forward them to router R2, which has the only working WAN link.

Figure 16.8 Problem: Extra Routing Hop When the R1 WAN Link Fails

A better plan links the HSRP role to the WAN link state. For instance, with interface tracking and preemption, you can configure HSRP as follows:

If R1’s WAN link is up, make R1 HSRP active.

If R1’s WAN link is down, make R2 HSRP active.

To do so, HSRP object tracking on R1 monitors the state of the R1 WAN interface. If the interface fails, HSRP lowers R1’s HSRP priority, allowing R2 to preempt R1. For example:

1. R1 notices its WAN interface fails.

2. R1 lowers its HSRP priority by 20 points.

3. HSRP messages allow both R1 and R2 to realize that R2 has a better (higher) priority and has the right to preempt.

4. R2, using preemption, becomes the HSRP active router.

As a result, while router R1’s WAN link is down, packets from the hosts on the left flow first to R2 and then out a WAN link. Later, when R1’s WAN link recovers, it preempts router R2 and becomes active again.

HSRP Recovery and Preemption

The HSRP preemption also dictates what happens when a formerly active router recovers, independent of any HSRP tracking. HSRP disables preemption by default but supports it on any router.

First, consider the following scenario:

1. R1 has priority 110, and R2 has priority 100, so R1 wins and becomes active, while R2 becomes standby.

2. Later, R1 fails, so R2 becomes active.

3. When R1 recovers, what happens?

With the default setting of no preemption, R2 remains the active router. With only two routers in the HSRP group, when R1 recovers, it moves to a standby state, ready to take over from R2 when R2 next fails. In that case, the operations staff can choose when to make the failover happen, for instance, off-shift, when a minor outage has no impact.

If you prefer, you can enable preemption. With preemption, in this case, or any case in which a new router appears in the group that has a better (higher) priority than the active router, it takes over as the active router. The switch to a new active router is methodical but quick, without waiting for timers like the Hello and Hold timers to expire, but it might cause a disruption in packet flow for a second or two. (Note that preemption does not apply to cases where the priorities tie, but only when the new router has a higher priority.)

HSRP Versions

Cisco routers and Layer 3 switches support two versions of HSRP: versions 1 and 2. The versions have enough differences, like multicast IP addresses used and message formats, so routers in the same HSRP group must use the same version. Suppose two routers configured in the same HSRP group mistakenly use different versions. In that case, they will not understand each other and will ignore each other for the purposes of HSRP.

There are good reasons to use the more recent HSRP version 2 (HSRPv2). HSRPv2 added IPv6 support. It also supports faster convergence when changes happen using shorter Hello and Hold timers, while HSRPv1 typically had a minimum of a 1-second Hello timer. Table 16-3 lists the differences between HSRPv1 and HSRPv2.

Table 16-3 HSRPv1 Versus HSRPv2

Ensure you understand how a router chooses the virtual MAC shown in the table. HSRPv1 supports 256 groups per interface, while HSRPv2 supports 4096. You can represent decimal values 0..255 with two-digit hexadecimal equivalents of 00..FF, while decimal values 0..4095 require three hex digits from 000..FFF. The HSRP virtual MAC addresses use the hex equivalents of the configured decimal HSRP group number as the last two or three digits of the virtual MAC address as follows:

HSRPv1: 0000.0C07.ACxx, where xx is the hex group number

HSRPv2: 0000.0C9F.Fxxx, where xxx is the hex group number

For example, an HSRPv1 group 1 would use virtual MAC address 0000.0C07.AC01, while an HSRPv2 group would use 0000.0C9F.F001. For group decimal 100 (hex 64), they would use 0000.0C07.AC64 and 0000.0C9F.F064, respectively.