Catastrophic zigbee2MQTT failure

I just checked and I have an install of HA that has now been running for 375 days without a reboot. This includes a Conbee II configured via Zigbee2Mqtt handling 45 devices. Yes, it is running an old version of HA OS (12.4) but if it ain’t broke don’t fix it especially as it is not my primary installation.

I’ve used a variety of naming conventions and now tailor the convention to help me identify where things are and what they are. Although with the way HA works sometimes it is tricky as you might have a temp entity in one location but the device in another location.

And there is the nub of it.

Don’t do any updates……

I’ll get this up and running and lock it down, thanks for the insight :slight_smile:
Now need to work out how to stop Home Assistant showing the updates.

I now have 84 out of the 120 paired again with 9 devices paired but now declared unsupported and another 10 or so not reporting for over 14 hours.

It’s not just that zigbee2MQTT suddenly declares it has seen itself in a mirror and panics it is all the pain at getting it going again.

By design you get a 4 minute window, there is a workaround, it it the whackamole method of pairing, you set the device off to pair only to find it is declared unsupported, requiring another ladder climb, or dive under a cabinet, or visit to the sheds in the rain, to do another pair attempt.

If zigbee2MQTT drops all your devices, you aren’t pairing 120 times it is nearer a 140 pairing attempts.

The only redeeming factor here is that zigbee2MQTT while forgetting every device suddenly remembers them when repaired.

I am migrating as much as I can onto a reliable zigbee platform that I have had no issues with for 10 years.

As much as I can I will replace with Shelly or Eufy leaving me with a minimal core of mains powered zigbee devices.

So when zigbee2MQTT drops its pants again the effort and time spent restoring it will be reduced.

Nah sooo many devices either going offline, no longer supported by zigbee2MQTT or just plain unresponsive.

Have reduced devices down from ambitious 120 to a more manageable 40 devices not easily replaced.

I have moved as much as possible over to the reliable Hue zigbee, replacing no Hue compatible lighting with Hue replacements.

Replaced unreliable zigbee2MQTT door/window/gate sensors and leak sensors with the reliable eufy sensors.

There has been a suggestion that zigbee2MQTT outside of Home Assistant is more reliable is that something Home Assistant supports?

By reducing the number of devices on zigbee2MQTT is substantially reduces the effort/time taken to rebuild after zigbee2MQTT curls up its toes and spits the dummy. So progressing reliability.

What devices are these?

Yes you can run it outside HA. I don’t see that it will give you more reliability though.

Aqara leak Sensors
Aqara door/window sensors
Third Reality Plant sensors
Tuya USB switches
Aqara Blind controllers
Sonoff Presence sensors

to name a few, were working two days ago - zigbee2MQTT had a fit dumped all 120 devices, but kept them in the config file.

Most did not put reconnect, so re-paired them.

Ones I had not reached auto came back up in zigbee2MQTT but reported they were now unsupported.

Playing device wakamole right now trying the keep the minimal subset running, and migrating everything else on to reliable systems. I get something back on and working in zigbee2MQTT by the end of the day it is unresponsive. Switches that worked not longer report button activity.

I have heard by people who run zigbee2MQTT in a vm that they don’t have these issues and it is related to the Home Assistant implementation - I have seen this on zigbee2MQTT forums too, an issue with zigbee2MQTT under Home Assistant is not the same thing as an issue with zigbee2MQTT

I have several of these, they have not been dropped from support. Seems you have some underlying issues with your installation or hardware if they are listed as not suppported.

It did really well, for the first year it was crashing and losing devices all the time, every day I was re-pairing 10-15% of the devices.

That first year was miserable and involved many many zigbee2MQTT and Home Assistant rebuilds.

Since November when I rebuilt Home Assistant again, for the 4th time on a third different platform it stabilised.

Except where it died and everything went offline every 14-20 days easily fixed by a Home Assistant reboot and a coordinator power cycle. Everything would come back online and start working again, magic.

This time I noticed nothing had checked in with zigbee2MQTT for a 2-3 hours - unusual for the mains powered devices, this time it just dumped every device, claiming a conflict with another (imagined) coordinator with the same PanID, forcing a GENERATE using the same extended PANID and Network ID.

10% of the devices found there way back unaided but they were mostly coming back as unsupported devices - but it did prove that the extended panid and networkID had not changed.

zigbee2MQTT had all the devices in its config file so it knew about them.

It seems the flakey behaviour exhibited throughout 2024 has returned.

Anyway I am committed to certain zigbee devices so will keep a minimal zigbee2MQTT running while I investigate another reliable zigbee host for them.

Noted and appreciate your responses,

these unsupported devices will eventually be supported if you keep re-pairing them. Eventually they are suddenly supported again.

But not reliable.

Really don’t understand it. I have been using Zigbee2MQTT since mid/end 2018, gone through several coordinators (always running an npm/node js installation) and it has crashed really never. Sometimes ( once or twice a year) I have to reboot the coordinator, but most of the time that is my own fault, and everything always comes back up.

I can get it down to 21 mains powered router devices and 26 battery devices.

I am replacing all the aqara devices Door sensors/leak sensors with Eufy ones.

The zigbee lights and switches replacing with Hue lights/switches.

I am replacing the zigbee power plugs with wave ones - I have a pile and Home Assistant wave has been rock solid for two years without issues.

The 21 double sockets are difficult to move right now, but they seem to be doing well.

The 9 agara blind controllers will take a while to replace, same as the sonoff presence sensors, but I can probably get away with Hue motion sensors.

That will just leave the plant sensors and watering modules.

SO reduced th elating on zigbee2MQTT form 120 devices to about 48, if nothing else will reduce the re-pairing time when it goes down again.

I also have had similar success. My only issues have been self-inflicted when testing the available integration methods across multiple hardware platforms.

I am envious! :slight_smile:

SO lets get this straight 90% of my Home Assistant devices are stable and working great!

Appletv
Awtric Pixel displays
Battery Notes
Dyson devices
The ESPHome device ePaper is a little flakey
LG TV
Govee devices
Harmony remotes
MG SAIC
Myenergi
Nanoleaf
Octopus
Plex media Server
Philips Hue (100+ devices) using zigbee NEVER needs any attention in over 10 years of use)
Shelly devices (although too slow for motion controlled lighting)
Powercalc
Sonoff wit devices
Tado - great - runs heating and hot water for every room in the house, doesn’t need any interaction ever
Victron
Yale locks
zwave

The only one that is unreliable and needs the most attention is zigbee2MQTT

As a hobby it is great and I have great hope it will stabilise but it is just too high maintenance for mainstream, so I need reduce my dependency on it.

Are you running those 100 devices via the Hue integration and parallel to other ZigBee devices installed via Z2M?

I have Aquara door/window sensors at all windows in the house and I use them even as generic binary sensors (soldered wires instead of the reed contact) as they are cheap AND reliable to me.

I’m not running more than 100 zinger devices, but I’m only missing a few until I reach this goal.
So, zigbee devices are the main part of my smart home devices.
Also powered by Zigbee2MQTT as HA addon.

Coordinator is a SonoffCC2652P.

I have two hue hubs each running ~50 devices each and I did have running, pretty reliably apart from outages every 14-20 days another 120 devices.

Once every 14-20 days a restart of Home Assistant and the coordinator would result in all devices coming back up and running.

Which was a HUGE leap in reliability from the re-pair 10-15% of devices everyday that was happening all last year.

So pretty happy with it until it dumped all 120 devices and has not returned to being really flakey again so massively reducing its workload.

Everything vanilla, everything has been factory rebuilt several times over the last 20 months on vis, Home Assistant green and the latest version from end of 2024 on its own micro pc. Gone through a range of coordinators as suggested by the people on this forum.

Everything else is solid and reliable, just zigbee2MQTT not a great loss, I’lll reduce the loading on it to under 50 devices that I cannot easily replicate on the more reliable services.

I’ll keep all the zigbee devices and revisit zigbee2MQTT in another year or so and try again.

But for now I need it reliable.

Really excited to hear it is reliable for others.

I may even run up the Home Assistant green and just use it for the mqtt and zigbee2MQTT then access through node red, that would be a great test everything vanilla no config on Home Assistant hardware :slight_smile:

Will do that once the new Hue Hub Pro comes out and I can put all the hue devices on one hub and free up a spare zigbee channel that does not clash with everything. That is due out in a week or so hopefully won’t be too long a wait.

Are you running different channels for the three ZigBee meshes you have? The Hue Hubs allow you change the channel. They are set to 25 by default.

In my opinion, it’s not Z2M. I have well over 100 ZigBee devices(mostly Hue devices), running from a single coordinator.

Yeh I have 20+ aqara door/window sensors they are great!

I have replaced the indoor ones with eufy door/window sensors and the batteries last over 2 years and they are solid, online 7x24. I’ll get another 8 for the outdoor sensors (in weatherproof containers) with eufy ones too.

The issue is not Aqara or zigbee, I have over 100 devices running on zigbee for over 10 years without missing a beat, but being Hue limited in variety and expensive. But appreciating their reliability, only had one device fail in ten years and Philips replaced it.

I started the Home Assistant zigbee2MQTT journey with the Sonoff CC2652P gone through a range of coordinators recommended by Home Assistant users.

I am happy that there is evidence that it can run reliably and hopeful that will be extended out further.

It is not without time/effort and money that this not working, it must just be my house is incompatible with zigbee2MQTT but is compatible with zigbee, no idea what I could do more to get it working.

But will keep zigbee2MQTT going just massively reduce the devices and my dependance on it and once I have cleared up another non clashing zigbee channel using the hue hub pro, due out in a week, then I can run up a zigbee2MQTT on a vanilla Home Assistant green and migrate the existing zigbee2MQTT devices across to see if it stabilises.

And access it from the main Home Assistant using node red.

You are actually hurting your overall stability by running 3 small and separate meshes.

Convert your spare dongles to routers. Sonoff’s web flasher is a breeze.

Out of curiosity, how many of your devices are router devices (mains powered) and how many are battery powered?