Backhaul faults that look like wireless

Why it happens

A client cannot tell a broken wire behind the access point from a bad radio link. Both look like silence. When the uplink fails, every client on that access point drops at once, with strong signal and clean air.

Switching faults start with spanning tree. One switch vendor’s walkthrough shows classic 802.1D isolating part of a network for 30 seconds¹ while it waits out twice the forward delay. Rapid spanning tree recovers faster, and ports configured as edge ports do not generate topology changes when the link toggles. An access point port left as an ordinary spanning tree port does, and each topology change makes switches flush the MAC addresses they learned. Smaller faults hide better. A duplex mismatch produces a high rate of transmission errors. A controller tunnel that pushes a frame past the path MTU must fragment it, and firewalls and other middleboxes tend to drop IP fragments, so pings pass while large frames vanish.

Power fails less cleanly. 802.3af delivers 12.95 W to the device and 802.3at delivers 25.5 W². An access point that gets less than it needs keeps running, badly. One vendor warns its performance will be unpredictable; another drops into a low-power mode that shuts off some transmit streams, for example when LLDP, the protocol switches use to negotiate power, is off on the port. When a switch’s budget runs out, port priority decides who keeps power, not plug-in order.

Wireless uplinks add their own failure modes. A mesh access point that relays backhaul on one radio roughly halves throughput at every hop, which is why one vendor recommends three to four hops at most. When a mesh access point loses its parent, one platform takes from 4 seconds on its fastest setting to 21 seconds on its standard one³ just to detect the loss, and every client below it waits. A design guide for EtherNet/IP over wireless counts network convergence beside roaming among the events that cause significant delay and packet loss, and calls for redundant infrastructure behind critical cells. Check the wire before you retune the radio.

How to confirm it

1. Line up drop timestamps with the access point's uptime, the switch port's link log, and the switch's PoE event log.
2. Check whether all clients on the access point dropped together. A radio problem picks clients off one at a time; a backhaul problem takes them all at once.
3. Read the switch port's error counters, and compare the speed and duplex the switch negotiated with what the access point reports.
4. Check the PoE class and power the access point negotiated against what it needs for full operation, and look for low-power or reduced-function warnings on the management platform.
5. Look for spanning tree topology changes at the drop times, and check that access point ports are configured as edge ports.
6. For mesh access points, check parent changes, hop count, and backhaul signal quality in the controller log.
7. Send large packets with fragmentation disallowed across the access point's path to find the real path MTU.

Ben Rutter. "Backhaul faults that look like wireless." OT Wireless, published October 5, 2026. https://otwireless.com/causes/backhaul-faults-that-look-like-wireless/

Rutter, B. (2026, October 5). Backhaul faults that look like wireless. OT Wireless. https://otwireless.com/causes/backhaul-faults-that-look-like-wireless/

@misc{rutter2026backhaulfaultsthatlooklikewireless,
  author = {Rutter, Ben},
  title = {{Backhaul faults that look like wireless}},
  year = {2026},
  howpublished = {\url{https://otwireless.com/causes/backhaul-faults-that-look-like-wireless/}},
  organization = {OT Wireless},
}