VRRP vs HSRP vs GLBP: First Hop Redundancy
Every host has one default gateway, and that gateway is a single point of failure until you make it virtual. Compare VRRP, HSRP, and GLBP with real configs, timers, and failure checks.

On this page
- At a Glance: VRRP vs HSRP vs GLBP
- The Default Gateway Is a Single Point of Failure
- VRRP: The Multivendor Default
- HSRP: The Cisco Default
- GLBP: Load Sharing With One Virtual IP
- Choose by the Constraint, Not by the Acronym
- Timers, Preemption, and Tracking
- Failure Modes and Verification
- Document the Virtual Gateway
A host with one default gateway has one path off its subnet. The router that owns that address can fail, lose its uplink, or be rebooted for an upgrade, and every host on the VLAN loses connectivity even though the rest of the network is healthy. First hop redundancy protocols solve that by giving the hosts a virtual address and a virtual MAC that two or more routers can share.
The three names that come up are VRRP, HSRP, and GLBP. They look similar in a diagram and behave differently in an incident. VRRP is the open standard, HSRP is the Cisco default, and GLBP adds per host load sharing that neither of the other two provides. This guide covers how each one elects a forwarder, how to configure and verify it, and which failures actually matter.
At a Glance: VRRP vs HSRP vs GLBP
| VRRP | HSRP | GLBP | Anycast gateway | |
|---|---|---|---|---|
| Standard | Open, RFC 5798 and RFC 9568 | Cisco proprietary | Cisco proprietary | EVPN/VXLAN design pattern |
| Vendor support | Broad | Cisco and some emulations | Cisco platforms that support it | Fabric platforms |
| Roles | Master and backup | Active and standby | AVG and up to four AVFs | Every leaf forwards |
| Virtual MAC | 0000.5e00.01XX |
v1 0000.0c07.acXX, v2 0000.0c9f.fXXX |
0007.b400.XXYY |
Shared anycast MAC |
| Load sharing | No, one master | No, one active | Yes, per host or weighted | Yes, local leaf |
| Preemption default | Enabled | Disabled | Disabled for the AVG | Not applicable |
| Group range | 1 to 255 | v1 0 to 255, v2 0 to 4095 | 0 to 1023 | Not applicable |
| Best fit | Multivendor pairs | Cisco only pairs | Cisco pairs that need active active | Large leaf and spine fabrics |
The Default Gateway Is a Single Point of Failure
The mechanism is simple enough to describe in one sentence: the routers agree on a virtual IP and a virtual MAC, one of them answers ARP for the virtual IP, and the hosts send their off subnet traffic to that MAC. When the active router fails, the standby takes over the virtual MAC and continues answering ARP, so the hosts do not need to change their configuration.
The virtual MAC is the part that makes failover transparent. A host that already has the virtual MAC in its ARP cache keeps sending frames to the same destination, and the switch learns that MAC on a different port after the failover. The host does not need a new ARP entry, a new route, or a new default gateway.
The terminology changes with the protocol, which is half the confusion. VRRP has a master and backups. HSRP has an active and a standby. GLBP has an Active Virtual Gateway that answers ARP and up to four Active Virtual Forwarders that carry traffic. The underlying idea is the same in all three: one virtual identity, more than one physical owner.
VRRP: The Multivendor Default
VRRP is the open standard, originally RFC 2338, then RFC 3768, then RFC 5798 for version 3, and now RFC 9568. It is the right choice when the pair is not from one vendor, or when the organization wants a protocol that any platform can implement.
The Cisco configuration is short. The virtual IP is the address the hosts use, and the priority decides the master. The interface address is the real address of the router.
interface Vlan10
ip address 10.10.10.2 255.255.255.0
vrrp 10 ip 10.10.10.1
vrrp 10 priority 110
vrrp 10 preempt
vrrp 10 track 1 decrement 20
The same group on the second router uses the lower priority and the same virtual IP.
interface Vlan10
ip address 10.10.10.3 255.255.255.0
vrrp 10 ip 10.10.10.1
vrrp 10 priority 100
vrrp 10 preempt
vrrp 10 track 1 decrement 20
Junos puts the VRRP group under the interface address.
set interfaces irb unit 10 family inet address 10.10.10.2/24 vrrp-group 10 virtual-address 10.10.10.1
set interfaces irb unit 10 family inet address 10.10.10.2/24 vrrp-group 10 priority 110
set interfaces irb unit 10 family inet address 10.10.10.2/24 vrrp-group 10 preempt
set interfaces irb unit 10 family inet address 10.10.10.2/24 vrrp-group 10 track interface ge-0/0/1 priority-cost 20
Linux hosts and appliances commonly use keepalived, which speaks VRRP.
vrrp_instance VI_10 {
state BACKUP
interface eth0
virtual_router_id 10
priority 110
advert_int 1
authentication {
auth_type PASS
auth_pass 4f3c2a
}
virtual_ipaddress {
10.10.10.1/24
}
}
Verification is the same idea on every platform: confirm the state, the master address, and the virtual MAC. On Cisco, show vrrp brief prints the group, priority, state, master address, and virtual IP in one row.
The virtual MAC for VRRP version 3 over IPv4 is 0000.5e00.01XX, where XX is the VRID in hex. Group 10 becomes 0000.5e00.010a. For IPv6 the range is 0000.5e00.02XX. Seeing that MAC in the switch table is a quick way to confirm which router currently owns the gateway.
On a Linux box, the VIP appears on the interface only while the node is master.
ip -br addr show eth0
tcpdump -ni eth0 vrrp
The capture should show advertisements every second. If it shows advertisements from two different physical MAC addresses with the same VRID, you have a split brain.
HSRP: The Cisco Default
HSRP is Cisco proprietary and very common in Cisco only environments. Version 1 supports groups 0 to 255 and uses the virtual MAC range 0000.0c07.acXX. Version 2 expands the group range to 0 to 4095, uses 0000.0c9f.fXXX, adds millisecond timers, and is required for IPv6. Use version 2 unless an old device forces version 1.
standby version 2
!
interface Vlan10
ip address 10.10.10.2 255.255.255.0
standby 10 ip 10.10.10.1
standby 10 priority 110
standby 10 preempt delay minimum 180
standby 10 track 1 decrement 20
The second router uses the same group and virtual IP with a lower priority, usually 100.
The show standby brief output is the one to keep in an incident runbook. It shows the group, the priority, whether preemption is configured, the current state, and the virtual IP in one row.
The virtual MAC for group 10 under version 2 is 0000.0c9f.f00a. Under version 1 it would be 0000.0c07.ac0a. The difference matters when you are tracing a MAC in the switch table or filtering a capture.
HSRP preemption is off by default, and that is not a bug: it means a router that reboots and comes back does not immediately take the active role away from the router that has been carrying traffic. If you enable preemption, add a delay so the recovering router has time to build its routing table before it takes the gateway.
GLBP: Load Sharing With One Virtual IP
GLBP is the Cisco protocol that solves the idle standby problem. In VRRP and HSRP, only one router forwards at a time. You can run multiple groups with different active routers and assign different gateways to different hosts, but that is manual and it multiplies the configuration. GLBP keeps one virtual IP and lets the Active Virtual Gateway hand out different virtual MACs to different hosts, so up to four Active Virtual Forwarders carry traffic at the same time.
interface Vlan10
ip address 10.10.10.2 255.255.255.0
glbp 10 ip 10.10.10.1
glbp 10 priority 110
glbp 10 preempt delay minimum 180
glbp 10 load-balancing host-dependent
glbp 10 weighting 100 lower 90 upper 95
glbp 10 weighting track 1 decrement 20
The lower and upper thresholds decide when the forwarder stops and resumes forwarding. The load balancing mode decides which forwarder a host gets. round-robin is the default and rotates new hosts across the forwarders. host-dependent keeps a host on the same forwarder for as long as it is cached. weighted uses the configured weight, which is useful when the routers do not have equal capacity.
show glbp brief
Interface Grp Fwd Pri State Address Active router Standby router
Vl10 10 - 110 Active 10.10.10.1 local 10.10.10.3
Vl10 10 1 - Active 0007.b400.0a01 local -
Vl10 10 2 - Listen 0007.b400.0a02 10.10.10.3 -
The GLBP virtual MAC is 0007.b400.XXYY, where XX is the group and YY is the forwarder. Group 10, forwarder 1 is 0007.b400.0a01. The AVG answers ARP and assigns the forwarders, and each forwarder owns its own virtual MAC. That is why the hosts can use one gateway IP and still send traffic to different routers.
GLBP is powerful and it is also the hardest of the three to troubleshoot. There are more roles, more timers, and more state to correlate. Use it when the load sharing is worth that cost, not simply because it exists.
Choose by the Constraint, Not by the Acronym
The protocol decision usually follows from three constraints: vendor mix, whether both routers must forward, and whether the platform supports the feature. Everything else is preference.
If the two routers are from different vendors, VRRP is the only portable answer. If they are both Cisco and one active router is enough, HSRP is the lower risk choice because more engineers have operated it and the tooling is mature. If the requirement is that both routers must forward for different hosts, GLBP is the built in option, but check platform support first because not every Cisco platform implements it.
In a large fabric, an anycast gateway is usually better than a first hop protocol. Every leaf switch has the same virtual IP and the same anycast MAC, and a host uses the leaf it is attached to. There is no standby router and no failover event for the host to notice. That design depends on an EVPN/VXLAN fabric, so it is not a replacement for a two router closet.
Timers, Preemption, and Tracking
The default timers are conservative for a reason. HSRP and GLBP use a 3 second hello and a 10 second hold. VRRP advertises every second and declares the master down after roughly 3 seconds plus a priority based skew. Those values survive a busy CPU and a lossy link. They also mean a failure can take 10 seconds to detect, which is an eternity for a voice call.
| Protocol | Hello or advertisement | Hold or master down | Notes |
|---|---|---|---|
| HSRP version 1 and 2 | 3 s | 10 s | Version 2 supports millisecond timers |
| VRRP version 2 and 3 | 1 s | About 3 s plus skew | Skew is based on priority |
| GLBP | 3 s | 10 s | Redirect and timeout timers also apply |
Lowering the timers is allowed and often necessary. It is also how people create outages. A 200 ms hello on a link that occasionally drops a frame will cause a failover that would not have happened at the default. Test the values under load, and keep them identical on both routers.
interface Vlan10
standby 10 timers msec 200 msec 750
HSRP version 2 accepts millisecond values. The hold time must be at least three times the hello time, and the platform enforces a minimum. VRRP sub second timers are platform dependent, so check the command reference rather than copying a value from another vendor.
Preemption is the other setting that causes surprise. HSRP and GLBP leave it off, so a recovered router stays in standby. VRRP leaves it on, so a recovered router with a higher priority takes the master role back. In both cases, a preempt delay is what prevents a router from taking traffic before its routing table is ready.
standby 10 preempt delay minimum 180
Tracking is what connects the gateway role to the health of the rest of the path. Tracking only the LAN interface is not enough, because the interface can stay up while the uplink is gone. Track an IP SLA probe or a route, and decrement the priority enough to force a failover.
ip sla 1
icmp-echo 203.0.113.1 source-interface GigabitEthernet0/0
frequency 5
ip sla schedule 1 life forever start-time now
track 1 ip sla 1 reachability
interface Vlan10
standby 10 track 1 decrement 20
With a priority of 110 and a decrement of 20, the router falls to 90 and the standby at 100 takes over. The numbers have to be chosen together. A decrement of 5 on a priority of 110 leaves the router at 105, which is still higher than 100 and produces no failover at all.
Failure Modes and Verification
The most common failure is not a router dying. It is the two routers losing sight of each other while both stay healthy. The result is a split brain: both think they are active, both answer ARP, and hosts get an inconsistent gateway MAC. The symptom is intermittent connectivity that looks like a switching loop but is not.
The first check is the protocol state on both routers.
show standby brief
show vrrp brief
show glbp brief
If both show Active or Master, check the Layer 2 path between them. The VLAN must be allowed on the trunk, the native VLAN must match, and no ACL can block the protocol. The multicast and port numbers are protocol specific, and a capture on the VLAN settles the question quickly.
| Protocol | Multicast | Transport |
|---|---|---|
| HSRP version 1 | 224.0.0.2 | UDP 1985 |
| HSRP version 2 | 224.0.0.102 | UDP 1985 |
| VRRP | 224.0.0.18 | IP protocol 112 |
| GLBP | 224.0.0.102 | UDP 3222 |
A capture on a Linux host or a mirrored port shows whether hellos are arriving. The filter for HSRP is udp port 1985, and for VRRP it is proto 112.
The second check is the MAC table. The virtual MAC should appear on exactly one physical port at a time. If it appears on two ports in the same VLAN, either a failover is in progress or both routers are forwarding. On Cisco, show mac address-table address 0000.5e00.010a answers that for a VRRP group 10 virtual MAC.
The third check is the path after the gateway. A router can own the virtual IP and still have no route to the destination. That is why tracking matters, and why a first hop protocol is not a substitute for a routing protocol.
Document the Virtual Gateway
A subnet record that lists 10.10.10.1 as the gateway is incomplete. The address is virtual, two routers can own it, and the group number and virtual MAC decide which one is active. Record the subnet, the virtual IP, the group, the virtual MAC, the priority, the preemption setting, the tracking objects, and the physical pair.
The subnet list is a good place to start because it shows which networks even have a virtual gateway. A gateway address that is a real interface address on one router is a single point of failure that nobody labeled.
curl --fail --silent --show-error -H "Authorization: Bearer ${OBELINF_API_KEY}" \
"https://api.obelinf.com/${OBELINF_ORG}/subnets?all=true" \
| jq -r '.subnets[] | select(.gateway != null) | [.cidr, .name, .gateway] | @tsv'
10.10.10.0/24 users 10.10.10.1
10.20.20.0/24 voice 10.20.20.1
10.30.30.0/24 management 10.30.30.1
The output does not tell you which of those gateways are virtual. That is the documentation gap to close. Add the group and the physical pair next to the subnet, and the next person can verify the failover without rediscovering the design.
Obelinf keeps subnets, VLANs, and device interfaces in one inventory, which makes it practical to record the virtual gateway beside the two routers that serve it. The VLAN and subnet management view is the natural place to keep that relationship visible.
The first step is to take the subnet list above and mark every gateway as physical or virtual. The physical ones are the single points of failure. The virtual ones need a group number, a virtual MAC, and a named pair. Either answer is useful, but only one of them survives the next router reboot.
Frequently Asked Questions
What is the difference between VRRP, HSRP, and GLBP?
Which first hop redundancy protocol should I use?
What are the default HSRP and VRRP timers?
Is HSRP preemption enabled by default?
Why are both routers active in the same HSRP or VRRP group?
Does first hop redundancy replace a routing protocol?
How should I document a virtual gateway?
Stop reaching for a spreadsheet
Obelinf keeps every subnet, device, circuit, and rack in one live source of truth, with audit logs and a topology view. Free for personal use.
Related Articles

DAC vs AOC vs Fiber: SFP Compatibility Traps
The port is usually fine. The coding, speed, FEC mode, or fiber grade is not. A practical guide to DAC, AOC, and optical transceivers, with the commands that show what the module actually is.
Read more
SNMP Explained: OIDs, Polling, and Real Data
SNMP is a database you query, not a monitoring system. Learn the OIDs that matter, why 32 bit counters lie, and how to turn a poll into a bandwidth number you can defend.
Read more
SNMP Traps vs Polling vs Streaming Telemetry
Polling, traps, and model driven telemetry describe the same switch but prove different things. Configure all three, then read each one for what it can actually confirm.
Read more