I am looking to created a high availability cluster so need to move my dongle to a machine that won’t change. I have node A & B and a pi acting as witness. The Pi will be the point in the network where the dongle plugs in. So if A fails over to B the pi will keep doing all its same jobs. As this is a single point of failure, I only want the pi to run the bridge and be where the dongle physically interfaces and have everything else on the HA cluster.
Currently I have the pi running zigbee2MQTT and I have it communicating with MQTT running in Home Assistant. I have confirmed this is working by publishing a topic to permit joining to the network in home assistant and the motion sensor I have for testing was able to connect and published it lux value back to mqtt in home assistant.
Now the question is, how do I manage the devices? I can see what the device is broadcasting but it’s not being shown as a device in home assistant. I was using zigbee2mqtt in home assistant before and that gave me a GUI button for permit join, nicer than trying to remember a command, and it passed the device nicely to home assistant for use in automations. Chat GPT said I could keep using zigbee2mqtt if configured correctly. But it doesn’t know how to configure it correctly and nor do I.
Should I be using zigbee2MQTT as the bridge between the dongle and MQTT and between MQTT and Home assistant? Or is there a simpler way to do it? Please could someone point me in the right direction.
I would move your MQTT broker to be on the Pi as well as Z2M (in docker). I would install MQTT as a standalone app for performance. You can still enable the GUI on Z2M running on the Pi and any other “essential” tools you might need (zwave2js, etc).
That way, should A fail over to B, you won’t have to migrate MQTT between A and B and only have 1 ip address to deal with and as far as HA knows, nothing changed.
Never listen to what ChatGPT says. It’s wrong 90% of the time when it comes to anything after 2021. If you need help configuring Z2M, that’s not a problem, but please avoid using ChatGPT for HA answers. You’ll only end up with wrong answers.
I was hoping you’d not say that but based on more thinks I have read, Building z2m and MQTT on the pi is what everyone says. It what GPT said until I strong armed it into my way of thinking.
I will give it a go and see how I get on. I will enable to GUI on Z2M first as that could be a simple solution however
I should give an update, I added the front end to the Pi and that was all I needed to keep the bridge on the Pi and have MQTT and HA on the cluster. I access the bridge’s UI via the Pi’s IP on port 8080, works a charm. yes it’s not in Home assistant but I have a network homepage and working DNS so it’s easy enough to get too.
I appreciate the advice but it seems to be running well as set up in both performance and recovery when failed over and if I start struggling with performance I know where to point my investigation gun.