An IP address conflict or a network loop: two mistakes that cut you off from the internet

Some faults take down one device. And then there are faults that take down the entire network at once - nobody has internet, printers and shared drives are gone, and the router sits there blinking as if nothing happened. The two most common causes of such a total outage are an IP address conflict and a network loop. Both are trivially easy to create and surprisingly hard to find.

IP address conflict: two devices on one address

Every device needs a unique IP address in the network. When two devices share one, the others do not know whom they are actually talking to - one answers now, the other a moment later. The result is intermittent, random outages: everything works, a minute later it does not, and then it does again.

How does a conflict arise? Almost always through a manually set address: someone once hard-coded an address into a printer, camera or PC from a range that the DHCP server has meanwhile assigned to someone else. The other classic is a device brought in with an address configured for a different network - a technician’s laptop or a box from a supplier.

The worst variant: a conflict with the gateway

While two printers fight over one address, those two printers suffer. But when something claims the gateway address - the router through which all internet traffic flows - the whole network is in trouble. Part of the outbound packets starts going to the false gateway, which has no idea what to do with them. The internet drops out for everyone at once, intermittently and unpredictably, while the router runs flawlessly.

We have dealt with exactly this scenario in practice: a device appeared in a network with a hard-coded address colliding with a key network element. The result was months of “mysterious” outages that only disappeared after a systematic readdressing - no hardware replacement would have fixed them.

A network loop: the cable that takes everything down

The second killer is even more mundane: one cable too many. Network switches send some messages to everyone at once (broadcasts). If a circular path exists in the network, a broadcast returns to where it came from, gets sent again, returns again… and within seconds the network drowns in its own messages. That is a broadcast storm.

And a circular path is created trivially:

• an extra cable between two switches (“just in case” or by mistake),
• both ends of a cable in the same switch - a tidying-up classic: a loose cable must belong somewhere,
• two wall sockets linked together that lead to the same switch in the cabinet,
• a second switch under a desk that nobody knows about, connected twice.

The difference from a conflict: a loop takes the network down instantly and completely. Nothing works, not even internal services, and the switch LEDs blink frantically, all ports at once. What the difference between a router, switch and hub actually is, we explained in a separate article.

How to tell which one you have

First aid during an outage

1. What changed last? A plugged cable, a new device, furniture being moved, cleaning. Total outages almost always have a trigger.
2. Unplug the most recently connected cable. If the network recovers within a minute, you have the culprit.
3. If you suspect a loop, disconnect switches and branches one by one until the outage disappears - that narrows the search to one branch.
4. If you suspect a conflict, disconnect the most recently added device. Which addresses live in your network can be checked following who is connected to my wifi.
5. Skip the router reboot as a fix - it does nothing for a loop, and for a conflict it merely postpones the next outage.

How to prevent it

Against conflicts, order in addressing helps: one DHCP server, reservations instead of manual addresses and a documented range. We covered it in detail in the article about a misconfigured DHCP server.

Against loops, technology helps: managed switches have STP (Spanning Tree Protocol) or simpler loop protection - they detect a loop and shut down the offending port before the storm breaks out. Cheap unmanaged switches usually lack this protection, so a network built on them is defenceless against a loop. Add labelled cables and a simple wiring diagram - during an incident it is the difference between minutes and hours of searching.

In a business it also pays to segment the network into smaller parts (VLANs): a storm or conflict then takes down one segment, not the entire operation.

How we see it

When we are called to a “mysteriously” failing network, an address conflict and a loop are the first things we rule out - and in a surprising number of cases also find. They have one thing in common: no purchase fixes them, only order does. Map the network, unify addressing, enable loop protection on the switches and document the wiring. From that moment on, outages stop being a mystery.

Conclusion

An IP address conflict and a network loop are two banal mistakes with a fatal effect: one is born from a manually set address, the other from one cable too many, and both can take down the entire network including the internet. You recognise them by the pattern (creeping outages vs. instant collapse), first aid is unplugging whatever changed last, and the permanent cure is called order: address reservations, managed switches with STP enabled and documentation.

This article is part of the computer networks overview.