core: KNX reaction is slow and unreliable

The problem

After update to 0.116.4 and updating config to work with my KNX setup I find it’s behavior is kinda degraded. Light devices are now switched on and off with visible delay, though before update it was almost instant, but the main problem that made me open this issue is lag of binary switches. HA doesn’t react to KNX button presses at all, until I switch something KNX related in HA interface. Then, it seems, that it establishes connection, and binary sensors work again, with visible lag though (1-2 seconds). After some time I see in logs Heartbeat to KNX bus failed. Reconnecting. And my binary switches doesn’t work again, until I switch something in the UI again. Tried restarting everything, doesn’t help. As I mentioned prior, it is a working setup, and before update nothing of described above was present. This is very annoying because I can’t switch third part of my lights, because they rely on HA to transmit commands between KNX and various platforms.

Environment

  • Home Assistant Core release with the issue: 0.116.4
  • Last working Home Assistant Core release (if known): can’t tell for sure something before breaking changes were introduced
  • Operating environment (OS/Container/Supervised/Core): Raspbian/Docker without supervisor
  • Integration causing this issue: KNX
  • Link to integration documentation on our website: https://www.home-assistant.io/integrations/knx/

Problem-relevant configuration.yaml

knx:
  tunneling:
    host: '192.168.10.55'
    # port: '3671'
    # local_ip: '192.168.10.80'
(commented out previously used option to see if this makes any difference (no it doesn't))
  binary_sensor:
    - name: Server room button
      state_address: '1/0/4'
      reset_after: 1000

Traceback/Error logs


Logger: xknx.log
Source: /usr/local/lib/python3.8/site-packages/xknx/io/tunnel.py:288
First occurred: 12:24:11 AM (1 occurrences)
Last logged: 12:24:11 AM
Heartbeat to KNX bus failed. Reconnecting. 

Additional information

Using Eibmarkt KNX EIB IP Interface, monitoring group bus in ETS, no problems.

About this issue

  • Original URL
  • State: closed
  • Created 4 years ago
  • Comments: 53 (28 by maintainers)

Most upvoted comments

Bug is still present in 0.118.1 Nothing changed at all.

@farmio

I will start to setup an environment to test 0.15.3, compile python 3.8 etc. Will report back soon

p.s thanks for helping us out with this issue. My HA install worked without any issues for more than to a year. Issues started a couple of weeks ago, but not sure from which version it started.

@farmio you’re right, apparently I can’t read 😃 Setting ignore_internal_state to flase helped with binary_sensor delay, thanks a lot!

Current documentation (your link) says default for ignore_internal_state is True and context_timeout is 1. These defaults will be changed in the linked PR.