MySensor gateway + Vera suddenly acting up (packet flooding)
I have a MySensor network with 4 nodes (3 window openers and 1 node to monitor a doorlock, rain, and temperature), using an Ethernet Arduino as a gateway, and the MySensors plugin on a Vera (UI5). This setup has worked flawlessly for about 6 months. No downtime, no resets required, no transmission or range issues, and instant response.
This week, one of the nodes stopped responding. As it turns out, that Arduino had somehow forgotten its sketch, and reverted to the default sketch (blinking light). I reloaded the correct sketch into the node, and after that it was working again. The node appears to have retained its settings, and no re-inclusion into the MS network was necessary.
Soon after that, everything stopped working correctly. Symptoms:
- Nodes no longer responding to commands from Vera.
- Vera no longer receiving temperature and status updates.
- Resetting the Gateway and Vera seems to fix things for a short while. Soon after a reset, the nodes will respond only intermittently and/or after a delay.
- After a reset, the Rcv led on the Gateway will sometimes blink for minutes on end. When reading the serial port I can see hundreds of packets coming in. During this flood, transmissions may or may not work.
All nodes and gateway running the 1.4 MS sketches.
It seems one node is responsible for the network flooding. It's the node with the rain sensor, having another node as parent. The sketch for that node already has flood control built in (rain sensors have a tendency to rapidly flip between states), and the Tx led on that node shows that the high level code is not actively making transmissions (not sure about the low level MS network libraries though). If I do make the node send an update (by opening / closing the lock for instance), the node shows a successful transmission.
Unplugging this node returns the rest of the network to a working state.
What could be causing this network flooding? As I've already established, there's nothing in my code that causes a multitude of packets to be sent. Could it be an issue with the mesh network? I've no idea how to diagnose the issue much less repair it.
A broken eeprom on one node cold mess upp routing?
A NRF-radio going crazy?
Hard to say... Not sure why it would result in flooding though.
After looking at the flood of packets more closely, it turns out it was a combination of factors: a faulty IR sensor in the lock caused rapid flipping of the sensor's state, and the flood control timeout was set to 0...