top of page

VRRP for Industrial Cellular Routers: Architecture and Hardware Selection Guide

52 minutes ago
9 min read
Two Wavetel industrial routers

VRRP is critical at any unmanned site where one router is the only way out. If that router fails, the whole site goes dark, and VRRP is what keeps the gateway available.

How much downtime can a remote site take before sending a technician costs more than a second router?

Take an unmanned roadside cabinet as the running example for this article. It is hypothetical, but the layout is common: a PLC watching cabinet temperature and power, an IP camera, and a managed switch tying them together, all reaching an operations center over cellular. One cellular router sits at the center. Multi-WAN and dual SIMs keep it connected when a carrier drops. They do nothing when the router itself dies. This article is about that second failure, how VRRP covers it, and how to pick hardware that does the job on cellular links.

What VRRP is and where it is used

VRRP (Virtual Router Redundancy Protocol, RFC 5798) lets two routers on the same LAN share one virtual gateway address. One router, the master, carries that address and forwards traffic. A backup router watches it and takes over the address if the master stops responding. The devices on the LAN keep pointing at the same gateway and need no change.

It is used where one router is the only way out of a site and its failure is costly: unmanned or hard-to-reach installations such as the roadside cabinet used as the example in this article, where sending a technician costs more than a second router.

The box is the single point of failure

A single industrial router can hold two modems and several SIMs, and fall back to a wired uplink. If Carrier A goes down, traffic moves to Carrier B. If an antenna lead snaps, the other modem takes over. For link failures, that works.

It does not help when the router fails: a power regulator burns out, flash corrupts after a bad power cycle, a remote firmware update hangs, a surge kills an Ethernet port, or someone servicing the next terminal block pulls the wrong cable.

In our cabinet, the PLC, camera and switch all use 192.168.10.1 as their default gateway. When that one chassis goes cold, the address goes with it. The switch still passes local traffic, but nothing leaves the cabinet. Both cellular networks may be fine outside the enclosure, and the site is dark.

Link failover inside one router is a separate problem, covered in our guide to failover vs load balancing vs bonding for 5G and Starlink. Surviving a dead router takes a second one.

How VRRP works on the cabinet LAN

In the cabinet, both routers carry the virtual IP (VIP), here 192.168.10.1, plus a shared virtual MAC. The PLC and camera keep pointing at the VIP and never learn which physical box is behind it.

VRRP pair at a remote cabinet

The master, the router with the higher priority, owns the VIP and sends advertisements to multicast address 224.0.0.18, once per second by default. The backup listens. If it hears nothing for roughly three advertisement intervals, it promotes itself, takes the VIP and sends a gratuitous ARP so the switch relearns which port leads to the gateway. The LAN devices change nothing.

Points to note

VRRP solves one problem, a failed gateway. It does not cover everything that can go wrong on a cellular site. These are the limits to plan around.

What the virtual IP does and does not carry

Three limits matter for cellular sites:

  • The VIP lives on the LAN only. VRRP knows nothing about the cellular side.

  • Carrier addresses do not move. Router 1 holds its address from Carrier A, Router 2 holds a different one from Carrier B. After a switchover, the site appears on the internet from a different address.

  • Session state does not follow. NAT and connection tracking tables are not synchronized, so open TCP sessions reset and reconnect through the new master.

Anything that reaches into the site, such as management access, SCADA polling or VPN peers, has to cope with that address change.

The cellular trap: LAN up, service down

VRRP only watches the LAN side. Suppose a backhaul fault takes Carrier A offline. Router 1 is still powered and its Ethernet link to the switch is up, so it keeps sending advertisements. Router 2 hears them and stays in standby. Meanwhile the PLC and camera send everything to Router 1, which has no working route out. The site is black-holed, and a healthy cellular link on Router 2 sits idle a few centimeters away.

The fix is WAN tracking. The master checks its own upstream path, and when that check fails it lowers its VRRP priority below the backup's, or gives up the master role. Router 2 then takes the VIP.

Normal operation and failover with WAN tracking

What the tracking tests matters as much as having it. On cellular, "link up" does not mean "service up". A modem can be attached to a tower and hold an IP address while the carrier's core network drops your traffic. Tracking that only looks at modem registration will call that router healthy. The probe should go beyond the first hop to a target you trust, such as your VPN concentrator, your own server or a public DNS resolver, using ICMP or DNS queries.

The return trap: preemption

The reverse problem shows up when the failed router comes back. By default VRRP preempts: Router 1, with priority 110, takes the VIP back as soon as it is up.

Say Router 1 reboots after a power dip while Router 2 carries traffic over Carrier B. Router 1's CPU and Ethernet ports are ready long before its modem has found a tower, attached, received an address and brought its VPN back. If it preempts immediately, it holds the gateway while it has no working uplink. The camera and PLC drop a second time, and this outage is self-inflicted.

A preemption delay timer fixes this by making the returning router wait before it reclaims the VIP. Don't take a generic number from a manual. Measure it on your own units: cold boot, cellular attach and tunnel up, with your SIMs, in your enclosure. Set the delay above that time, with margin.

Flapping is the related risk. If marginal reception makes a link drop and recover repeatedly, the pair can trade the VIP back and forth, resetting sessions each time. Some form of hold-down, where the primary must stay healthy for a while before it can take over again, keeps that in check.

Tunnels do not fail over with the gateway

Moving the VIP does not move the VPN. In our cabinet, PLC data and camera video ride an outbound tunnel to a central concentrator. When Router 2 takes over, it cannot inherit Router 1's tunnel. It has to build its own from the Carrier B address. Real recovery time is switchover time, plus tunnel re-establishment, plus the time for applications to reconnect. The VRRP part is usually the shortest.

How the tunnel behaves decides the rest:

  • IPsec: When Router 1 dies without saying goodbye, the hub keeps the old tunnel until its dead peer detection (DPD) timers expire. Conservative timers leave the hub holding a dead peer while Router 2 is already trying to connect. Match DPD settings on both ends to how stable your cellular links really are.

  • WireGuard: There is no separate negotiation to repeat. Once Router 2 sends an authenticated packet from its new address, a hub configured to accept a changed peer endpoint can start answering it. That usually leaves less to redo than IPsec or OpenVPN, though you should test it with your own hub.

  • Inbound access: Most SIMs sit behind carrier-grade NAT, so reaching a cabinet from outside needs a public or static IP plan, a private APN or dynamic DNS. DDNS adds propagation delay after a switchover. An outbound persistent tunnel to a hub avoids the problem, because the cabinet always dials in.

Measure tunnel recovery in staging before you ship pairs to the field.

Shared single points of failure

A second router also does not remove every single point of failure. The shared switch and the DC supply are still single points. If both routers hang off one supply or one breaker, one fault takes down the pair.

Hardware selection

Symmetric or asymmetric pair

Symmetric pair. Two identical routers. You stock one spare model, reuse one configuration template and get the same capacity whichever unit is master. Put the two units on different carriers and two dual-modem routers give you four independent cellular paths across two chassis. Because both can carry the full load, you can run without preemption and avoid a second gateway change after every recovery. Keep preemption, with a delay, if you want a specific unit to be master, for example because its antennas are better placed.

Asymmetric pair. A higher-capacity primary with a lighter backup, such as a 5G unit backed by a 4G unit. The backup may not carry everything, so decide up front what gets dropped during failover, for instance camera streams, and whether your firmware's policy routing or QoS can enforce it. Here preemption with a verified delay is the sensible choice, so traffic returns to the primary once it is truly working.

Where WR6 fits

A "VRRP" line in a feature list is not the answer by itself. What matters is the behavior around it, and the WR6 firmware covers the two behaviors this article turns on. It supports WAN link tracking with VRRP priority decrement: when the master's WAN link fails, the master lowers its own priority, so a backup that still has a working uplink takes over the VIP even though the master's LAN side is healthy. It also supports preemption, so a recovered unit can reclaim the master role. Wavetel confirmed both for this article; the public datasheets list VRRP in the routing stack but do not describe them.

Failover and recovery time on a WR6 pair is not a fixed figure. It is set by your configuration: the VRRP advertisement interval, the probe interval, timeout and attempt count used for WAN checks, and the preemption delay. Configure these for your site, then measure the result as described above. Two details are worth confirming against the manual for your firmware build before you order, because they decide how closely the pair matches the design: whether the tracking decision uses link state only or also probes beyond the first hop, and whether the preemption delay can be set long enough for your measured boot, attach and tunnel time.

The four WR6 models share one platform: 1×2.5GE, 4×GE and 1×SFP; Wi-Fi 6 AX3000; RS232 and RS485; 11 digital inputs, 2 digital outputs, 1 analog input, 2 relay outputs and an ignition input; GNSS; a four-pin Molex DC power connector or active PoE PD input; and an aluminium IP30 housing for desktop, wall or DIN-rail mounting. The datasheets give a -40 °C to 75 °C operating range. The routing stack lists static and policy routes, BGP, OSPF, RIP, NHRP and VRRP, alongside Multi-WAN, WAN failover and load balancing, with IPsec and OpenVPN. WireGuard is listed on the product pages and in the WR677, WR677-M and WR647-D datasheets, but not in the WR677-D datasheet, so confirm it there if your design depends on it. Management is by web GUI, SSH, TR-069, SNMP and RMS.

Model

Modems

SIM slots

Where it fits in a pair

Dual 4G, Active-Active

4

Both units 4G on different carriers, or the 4G standby behind a 5G primary

Single 5G (Sub-6 GHz, NSA/SA, falls back to 4G/3G) + ETH WAN

2

One cellular path plus wired uplink; put the partner unit on a different carrier

5G + 4G, Active-Active

4

Both units 5G+4G on different carriers, or either role in a mixed pair

Dual 5G (Sub-6 GHz, NSA/SA, falls back to 4G/3G), Active-Active

4

A matched dual-5G pair, or a 5G primary behind which a lighter unit stands by

Active-Active here means both modems can be in use at once. The WR6 product pages state that it does not combine the links into one bonded connection. Specifications come from the Wavetel v2.0 datasheets and product pages; confirm current availability with Wavetel.

Recommended pairs by scenario

Use the table as a starting point. Match the pair to what the site must keep running after a failure.

Scenario

Suggested pair

Why

The site must keep everything running, including video, after a failure

Two WR677-D, or two WR677-M, each router on a different carrier

Either unit can carry the whole site, so a failover does not reduce capacity

The site can shed video and keep only PLC data during an outage

WR677 as primary, WR647-D as standby

A lighter standby; decide in advance what is dropped

4G coverage is enough for the site

Two WR647-D, each router on a different carrier

Both units are 4G with two modems each, on separate carriers

A wired uplink is available alongside one cellular path

WR677, with the partner unit on a different carrier

One cellular path plus a wired uplink on each unit

The pair mixes 5G and 4G

WR677-M in either role

5G + 4G on one unit, usable as primary or standby

Treat every pairing in the table as a design to test on your own site, not as a configuration Wavetel has verified.

FAQ

Do TCP sessions survive a VRRP failover?

No. VRRP moves the gateway address and nothing else. Connection tracking and NAT state are not shared between routers, so open sessions reset and reconnect through the new master.

Does the VIP carry the carrier IP address to the backup router?

No. The VIP exists only on the cabinet LAN. Each router keeps its own carrier address, so VPN peers and any inbound access must tolerate a new address after a switchover.

Do I need preemption in a symmetric pair?

Usually not. If both units have the same capacity, staying on the current master avoids an extra gateway change. Keep preemption, with a delay, only if you want a particular unit to be master.

Can each router still use its own dual-modem failover?

Yes. Switching modems or SIMs inside a router is the first line of defense and does not move the VIP. VRRP is the second line, for when the whole box fails or loses every WAN path.

Recent Posts

See All
bottom of page