My network all of the sudden is now dumping TONs of these errors. They start after about 24 hours or so and are temporarily fixed by restarting the network.
They are all "Publish ‘get’ … delivery failed … " for all sorts of different parameters.
Any ideas?
For reference nothing touched any of the properties it is “getting”… so unclear why it seems to be getting every property for every device.
This behavior usually points to a memory leak, full message queue, or radio failure in the coordinator. It is common with setups like Zigbee2MQTT, Z-Wave JS, or Home Assistant.
When “Publish ‘get’ … delivery failed” floods the logs after a steady 24 hours, the system is attempting to poll device states, but the message buffer has overflowed or the connection to the adapter has stalled.
The most common causes and fixes include:
1. USB 3.0 Interference or RF Jamming If using a USB radio dongle (like a Sonoff or ConBee stick) plugged directly into the server, USB 3.0 ports emit interference that degrades signals over time. The coordinator retries failed transmissions until its message buffer fills completely.
2. Polling Loop in an Automation An automation, script, or custom polling integration may be requesting updates (like power consumption or temperature readings) too frequently. Over 24 hours, outbound requests stack up faster than the mesh network can deliver them.
3. Mesh Routing Loops or “Babbling” Devices A malfunctioning smart device on the mesh might be spamming corrupt packets or dropping off the network, causing route recalculation loops that exhaust the adapter’s available packet IDs.
4. Adapter Firmware Older coordinator firmware often has memory management bugs that cause the adapter to lock up after handling a specific volume of traffic.
Found it. Was a node-red module I downloaded that was supposed to help expose inovellis more easily. Turns out it was regularly calling GET on EVERY parameter on EVERY switch.