The last few days I’ve had a strange issue which has proven difficult to troubleshoot, however I think I finally found the answer. I wanted to share in case anyone else gets a similar issue. Basically majority of Zigbee devices became slow to respond, and some gave some really odd behaviour. My best guess is a change to HAOS may be the trigger. On the release notes there were some mention about additional wifi device support, this sparked me after a few days of head scratching to disable wifi in the BIOS which for now appears to have solved the issue.
Full details:
I’ve been running around 70 Zigbee devices with no issues for some time, everything is near instant.
My setup is native NUC running core 2021.9.4, Supervisor 2021.09.0, OS 6.2, Zigbee2MQTT 1.21.1-1, Mosquitto broker 6.0.1. ZZH CC2652R on USB extension, all on a NUC5i7RYH. I updated most of these in the last week from -.1 or -.2 versions. Things started to fail with one of these updates (but I was unable to pinpoint which)
I noticed the issue starting with a number of these buttons. According to the Z2M page they can give a hold action if the zigbee network is slow, rather than a single action. I noticed one button had this issue, and when checking all did. Adding delays in didn’t help reliably. Other buttons worked fine, however I think everything was running slowly, I just didn’t notice due to the stress of why the buttons were going crazy.
I don’t think I ever looked at the Z2M logs in debug mode, so I wasn’t sure what I was looking for, however I noticed a lot of messages every few seconds like Device ‘Front Door Light’ requested OTA. These repeated over and over. This seemed odd… I checked with an HA buddy and he did not seem to have similar messages, still the relevance escaped me.
I powered down every zigbee device in case I had a bad behaving router, powering back on and they seemed to work for a few minutes, I re-paired many, I paired closer to the NUC… Some things seemed to work occasionally, but none persisted long term. I restored snapshots and all the normal stuff.
It finally hit me, when I opened my under stairs cupboard, the light was taking around 5 seconds to trigger from a door sensor. Everything was slow, not just the buttons.
Having remembered about some wifi stuff in HAOS as well as the possible interference with Zigbee, (NUC is cabled, and Zigbee on USB extension), I disabled wifi within BIOS and now things have just started working, the buttons no longer dim things instead of turning on and off, sensors are back to near instant, and so on. There are no longer loads of messages on devices requesting OTA.
My best guess is somehow the on-board wifi (which I guess never worked), started working with new drivers or similar, and that it was scanning through the various frequencies, some which were perhaps interfering with the zigbee network. Perhaps this is why it worked sometimes and not others.
I am not 100% on this, and also I don’t know really how to report it as a possible bug, but here’s hoping this long post might help someone who has a similar setup. I know there is advice to not use wifi, however I was not using wifi - well at least to the best of my knowledge!
Fingers crossed
