BUFFER FULL, does anyone know what's the cause?

[Tee86]

What happened?

I have the latest HA, latest z2m, latest emqx, Dongle-P with latest FW, updated Debian.
3 days ago I got
failed with status '(0x11: BUFFER_FULL)' (expected '(0x00: SUCCESS)
on all my powered devices (a.k.a. routers) it happened every day since then, yesterday twice.

@Koenkk , do we know what's behind that error. I'm on the latest Z-stack FW for at least 3 months and never had this issue. I did not make any significant changes to my setup recently. Just added one IKEA TRETAKT plug, but there were already 3 before this one, never had issue with TRETAKT's.

I have currently 32 devices connected to coordinator. Is it too much? Is there any way to limit that amount? I can see via map that the most recent TRETAKT is not even directly connected to the dongle, it's just routed.

I'm gonna try tomorrow that 2022 firmware that was suggested in another post, although, it has a mixed feedback... I'd rather find out what the culprit is...

Funny thing is that somehow this affects only powered devices, end-devices remain online, even tho their routers reports to be offline. First time it happened I noticed it after 15 hours, all my 41 powered devices were offline, and all 47 of my end-devices were reporting still online

What did you expect to happen?

No response

How to reproduce it (minimal and precise)

No response

Zigbee2MQTT version

2.1.3

Adapter firmware version

20240710

Adapter

Sonoff Dongle-P

Setup

Docker installation, Pi

Debug log

No response

[Koenkk]

Could you try the fw from Koenkk/Z-Stack-firmware#518 (comment) ?

[Tee86]

Thank you @Koenkk . I downgraded the firmware before your reply to 20221226, I did that about 24 hours ago. I have had no "BUFFER FULL" so far.
Is there some significant improvement in 99250304 in comparison to 20221226?
I noticed that with 20221226 devices react better than with 2024 version I had originally.
The only downside I'm noticing is lower LQI - which might be caused by the mesh rebuilding, right?

@Koenkk , There is one thing I found out yesterday. I run this server for a client of mine, and he received this device a few days ago from his energy supplier. I confirmed that it communicates via Zigbee with the energy meter. Could that be related? Is it possible that it might disrupt the network and cause error like that?

[Koenkk]

I confirmed that it communicates via Zigbee with the energy meter. Could that be related?

Maybe it could cause interference, but it is hard to judge that.

Could you also give 99250304 a try?

[Tee86]

@Koenkk ,a few updates:
I ran 20221226 since Friday 11AM till Tuesday 11AM, so full 4 days. I had no "buffer full" errors whatsoever during that time, but as I mentioned, the LQI's were quite low, and I had several devices dropping out temporarily (mains powered), and one device (Sonoff door sensor - battery powered) went offline completely and had to be repaired.
I flashed 99250304 on Tuesday, no errors so far, but it's been just one day. I'll monitor it in upcoming days and will drop an update sometime next week.
Thanks for your help btw