My shed AP had been intermittently dropping off the Unifi controller for weeks. It surfaced during a controller migration when push notifications started firing: “device disconnected for 21 minutes, reconnected.” The other AP and switches were fine. Just the shed.

The Setup

The shed is about 0.1 miles from the house. A pair of NanoStation Loco M2 devices form a point-to-point wireless bridge between the two buildings. A UAP-AC-Lite hangs off the shed-side bridge to provide WiFi coverage out there.

1
2

[House] --ethernet-- [Loco M2 AP] ~~wireless~~ [Loco M2 Station] --ethernet-- [AC-Lite]
         .13                                          .14                       .137

The bridges run airOS (not Unifi), so they have their own management UIs separate from the Unifi controller.

Finding the Bridges

First challenge: I didn’t know the bridge IPs. They’re not managed by the Unifi controller and weren’t documented anywhere. A quick scan of the subnet found them at .13 and .14, each running airOS v6.3.11.

The Smoking Gun

The house-side bridge looked fine at 6% CPU. The shed-side bridge was at 100% CPU. After a reboot, it came right back up to 100%. Not stale state. Something was actively consuming all resources.

The logs told the story:

1
2
3
4

udapi-bridge[873]: unms: connecting to 13.249.141.93:443
udapi-bridge[873]: connection error (redacted.uisp.com:443): Timed out waiting SSL
udapi-bridge[873]: unms: connecting to 13.249.141.75:443
udapi-bridge[873]: connection error (redacted.uisp.com:443): Timed out waiting SSL

The UISP/UNMS agent was trying to connect to redacted.uisp.com every 20 seconds, timing out on SSL each time, and immediately retrying. On a device with the processing power of a calculator, that tight retry loop consumed everything.

Both bridges showed “UNMS: Enabled but unreachable” on their status pages. The cloud controller it was trying to reach either never existed or had been decommissioned long ago.

The Fix

On each bridge: Services tab, disable UISP/UNMS, save. The shed-side CPU dropped from 100% to 3% immediately. The AC-Lite reconnected to the Unifi controller within minutes.

What I Learned

The bridge link was never the problem. Signal strength was -29 dBm (excellent at 0.1 miles), CCQ was 97-99%, and the wireless link was negotiating 300/300 Mbps. The investigation plan assumed antenna drift or firmware issues, but the actual cause was a management agent running amok.

Check the logs before assuming hardware. The status page showed 100% CPU, and the logs showed exactly what was eating it. Could have saved time by going straight to the logs instead of trying reboots first.

Disable cloud agents you’re not using. UISP/UNMS was enabled on both bridges but pointed at a nonexistent cloud controller. The house-side bridge was handling it at 6% CPU (maybe fewer retries or faster timeouts), but the shed-side was drowning. If you’re not using a cloud management feature, turn it off.

Document your bridge IPs. These devices sat on the network for 200+ days without anyone knowing their IPs or checking on them. They’re now documented and the UISP agent is disabled on both.