Internet Failover Jamaica: Does Your Backup Link Actually Work?
Plenty of Jamaican offices pay two internet providers every month on the assumption that if one drops, the other takes over. Internet failover in Jamaica is sold as redundancy, billed as redundancy, and trusted as redundancy. The uncomfortable part is how often it has never worked and nobody knew, because nothing had forced it to try.
During one failover health check, we found a backup link that could not have carried a single production session. The route was there. The firewall knew the second provider existed. What was missing was a NAT rule on that link, so the moment traffic tried to leave through it, it would have been dropped on the way out. On paper the office had two providers. In practice it had one, plus a monthly bill for the illusion of a second.
Why "we have two providers" is not the same as failover
Two links give you redundancy only if the equipment in the middle is configured to move traffic from one to the other cleanly. That is a firewall and routing job, not an ISP job. FLOW and Digicel each deliver a working circuit to your door; what happens when one of them dies is entirely down to how your firewall is set up.
The common gaps we see in Jamaican offices are mundane and invisible until tested:
- The backup link has a default route but no outbound NAT, so return traffic never makes it back.
- Failover is configured, but the firewall never re-sources existing sessions, so everything that was connected hangs instead of reconnecting.
- Nobody has defined what "down" means, so a degraded-but-up link keeps being treated as healthy.
None of these show up on a bill or a status page. They show up during an outage, which is the worst possible moment to discover them.
Pulling the cable is not a real test
The most common "test" is unplugging the primary link and watching traffic move. It feels convincing and it proves almost nothing. Physically removing a cable is the one failure mode that is easy to detect, because the interface goes physically down and the firewall reacts instantly.
The outage you actually get in the real world is different and nastier: the link stays up, the light stays green, but the provider is dead somewhere upstream. Your firewall still sees a live interface and keeps sending traffic into a black hole. Unless it is running health checks that test reachability to something on the far side of the provider, not just link state, it will never fail over at all.
That is the scenario a cable-pull test never exercises, and it is the scenario that takes offices down.
Sessions that hang for 30 to 90 seconds
Even when failover does fire, there is a second trap. Existing connections were built using the primary link's public address. When traffic shifts to the backup, those sessions are still trying to use the old source address and simply stall. Staff see spinning tabs, dropped VoIP calls, and applications that time out, often for 30 to 90 seconds, before everything re-establishes on the new path.
A properly configured firewall re-sources active sessions onto the backup link so the gap is measured in a packet or two rather than a minute of dead air. The difference between those two experiences is entirely configuration. We have cut a real-world failover event down to a single lost packet with no production outage, purely by setting the firewall to behave correctly when the path changes.
What a real failover test looks like
If you want to know whether your redundancy is real, the test has to reproduce the realistic failure, not the convenient one:
- Confirm the backup link can actually carry traffic on its own, NAT included, before you ever need it.
- Simulate "link up, provider dead" by blocking reachability upstream, not by pulling a cable.
- Watch what happens to live sessions, not just new ones, and measure the gap.
- Fail back and measure that too, because a messy failback can be as disruptive as the original outage.
Most of this is a desk exercise plus a short, planned change window. The point is to find the broken NAT rule, the missing health check, or the hanging sessions on a quiet afternoon, rather than during a storm-season power event with the whole office watching.
Redundancy you have never tested is not redundancy. It is a second invoice.
Where this fits
A backup link is only one layer. Getting real value from two providers usually means steering traffic intelligently across both, which is what SD-WAN for Jamaican organisations does, and making sure the firewall underneath is not quietly undermining the whole design, which is covered in the firewall mistakes we see most often in Jamaican offices.
Free failover health check. If you pay for two providers and have never properly tested the second one, we will. Book a failover health check with Systems Rubix and find out whether your backup link would actually carry your business, before you need it to.