Zigbee Devices become unavailable - and loads of [zigpy.zcl] [0xb432:1:0xef00] Unknown cluster-specific command 36

have been using HA for a few month with some good success this year but started playing up recenty again.
Zigbee devices become unavailable over time or don’t become available when HA rebooted. I have a few WARNING (MainThread) [zigpy.zcl] [0xb432:1:0xef00] Unknown cluster-specific command 36 in my logs
Sometimes HA reports devices as seen very recently in the settings but they don’t respond or show wrong status.
It’s in a different room to Wifi NW and hard wired. System is Pi4 with Samsung Pro Endurance SD Card and I switched to MariaDB after initial problems

Zigbee Home Automation
ConBee II, s/n: DE2230813 - dresden elektronik ingenieurtechnik GmbH
60 devices and 81 entities

Any pointers would be helpful

Can you power cycle your ZigbeeDevices? I have seen similar and from my experience its almost always the device itself that is messing up.

Yes some I can and then it springs into action, sometimes for permanent sometimes only for one cycle and it stops again. Some devices I cannot power cycle like a sonoff mini switch as it’s hidden (still operatable manually but not power cyclable). Can MariaDB fail like the SQLLite does? That’s when I seen the problems before.

Another interesting example I found today. A motion sensor is shown unavailable in HA but an automation triggered by that same motion sensor and it also shows as last seen when the automation triggered. Must be a software thing.

Solution are several combined:
Change database to MariaDB
Chuck out SD card (SSD instead)
Lights are bad routers so more wired switches, plug in sockets and modules instead work better
Don’t add too many zigbee devices at once or change them around rooms too much.
Switch off 2.4G Wifi in main part of house

Hi @ChrisInUK - where would one change the DB to MariaDB - are you referring to Home Assistants recorder or the Zigbee integration - at the moment I am looking into a similar issue locally and running through similar troubleshooting steps.

Seeing a lot of [0x3907:1:0x0005] Unknown cluster-specific command 7 messages in my log file and would like to get to the bottom of them - this is for an IKEA Tradfri 5 button remote running the 0x23080631 firmware.

In supervisor add-ons you find MariaDB for home assistant. That provides a much more reliable history which seemed to have helped to some extend with my ZHA devices (Zigbee Home Automation). Also note that Lightbulbs don’t seem good routers. So I now use more switched sockets Ikea and Sonoff and 1 Tradfri repeater from IKEA and all client devices seem more stable now too. 99% reliability. Have a look here: https://whatarewefixing.today/240/fixing-home-assistant-database-problems-with-mariadb-addon/