Event caching issue

[jmachta]

Recently I was dealing with issue, which I am not sure, if it is a problem of z2m or any other element in the system. Hereby I am looking for a clue if anybody experienced this too. I will describe.

On Linux machine (based on Ubuntu) I have mosquitto, zigbee2mqtt and overall it is controlled by rules in openHab. Nothing really special. I use Conbee 2 zigbee dongle. In the system there are mainly Ikea Tradfri bulbs and switches, but the issue appears also with Aqara switch (I have just one). Now the problematic behavior.

When I press tradfri button (there is on and off event), the event shall be caught by z2m, published to Mosquitto, processed by openhab and then command to tradfri bulb shall be sent in opposite order. If event is "on", bulb turns on, obviously. However, sometimes the bulb does not come up upon "on" event (no message is received in z2m log), but if I press within short time "off" or "on" (in few seconds), the bulb comes up, "on" event from tradfri button appears. In z2m log I see the "on" event only. Then everything works fine again, on or off does what it is supposed to do.

From outside observation it seems, that zigbee message gets somewhere cached and gets processed by z2m only after next message is received. Does anybody have similar experience? Could this be issue of OS serial port cache?

1. Did anybody see similar behavior?
2. Did anybody try similar chain on Windows? I was thinking to convert to this awful OS in case it helped...

[jmachta]

I believe I can answer my question after a difficult weekend with my smart home. :-) Now the system is much more reliable (from reaction perspective), problematic behavior went away. There were basically following modifications to be done:

1. In OpenHab set the QoS in MQTT broker Thing object to 2, which enforces that each MQTT message is delivered exactly once. I also deleted QoS settings in MQTT channels, which I have set to 0 somewhere in the past.
2. I reworked all rules that were triggered based on z2m inputs. Previously I have triggered my actions based on item state update, I changed this to trigger based on item state change. And this probably made the miracle.

Explanation:
In log after pressing the switch I saw that button state has changed. But the response from OpenHab was quite chaotic, I saw several actions for the bulbs, not all of them made sense to me, e.g. on switch on, I saw commands for bulbs with brightness 0 etc...

My feeling is, that channel triggers were coming not only from action of the button (on, off) but also LQI value was considered an update to channel. Now the behavior is much more predictable.

I hope this helps someone... I was fighting with this issue for few months. It was annoying, but not annoying enough to force me to fix it. :-)

[jmachta]

After another year of playing with the system, I have to add, that the real solution was a dongle replacement. I originally used Conbee II, then I switched to Sonoff and all my issues have vanished. Hope this helps to someone.