I have an issue where the connection to my KNX network (Jung IPS 300 SREG, configured as IP interface) only works for a few seconds, and then stops working. It seems like it starts working only after I go to ETS5 and open the Bus tab to check the IP connection. After that HA will be able to connect and I can even turn on the light, but after about 10 seconds it stops working and I have to restart the KNX IP interface. However ETS5 can connect without problems even after it stops working in HA.
Here are the logs:
Maybe the “CONNECTIONSTATE_REQUEST” message is causing a problem?
Hi !
First of all, please check if your device is using the latest firmware and application version.
Is the device configured to use IP-Secure tunnelling?
This is used as heartbeat message to not automatically close the tunnel. But for any reason the tcp connection is already closed when sending this, so there might be some kind of networking problem here, but that’s just a guess (because there is no received KNX DisconnectRequest the specifications would require if it was a clean disconnect).
I’m not using IP Secure (I disabled it). I checked the version on the screen of the IPS and it says “2.000”, which seems to be the latest version. What did you mean by “application version”? I just did a full download to it from ETS5 a few days ago, imported device from from the catalog (version 1.0, which also seems to be the newest).
I checked with WireShark that ETS5 is also sending the connection state requests, but it also gets an OK response. It’s weird that while HA logs are showing “Could not establish connection to KNX/IP interface. ConnectionRefusedError: [Errno 111] Connect call failed”, ETS5 can still connect normally and everything there works fine. For HA to be able to connect, I have to physically turn off/on the IPS interface. Any idea what I could try?
Another thing I should add, I installed HA today on a VM in my TrueNAS. I used the KVM image (qcow2). The network interface is bridged to my LAN and obtains an IP via DHCP (192.168.1.45) and the IP of the IPS interface is 192.168.1.20. If I ssh into the HA (via the SSH addon), ip addr will display a different address (172.30.33.0/23), but I can ping 192.168.1.45 from my PC and it will succeed.
Also, since it is able to connect sometimes, I’m guessing the networking config is ok for the VM/HA.
If you SSH to the SSH-Addon you land in a Docker container (all Addons are) that’s not on host network - therefore the 172.x IP.
Well, something seems to be wrong in the network stack - it would work otherwise and not raise [Errno 104] Connection reset by peer.
Not sure if there is anything we can do to narrow down this issue from HA side. I guess something like Wireshark would be the better tool to find out what is going on there.
Another way would be to use telnet to read out the interfaces logs… but I don’t think these are really documented so you’d need to contact Enertex support. I’d try HA from a different machine before (a small test env would be enough like https://developers.home-assistant.io/docs/development_environment)
@farmio Thanks! In the end it was my stupidity… I had another device with the same IP in the network, I didn’t realize it because it was not using DHCP and that’s the only place I checked.
Interestingly, ETS had no problem with connecting to the KNX IP interface every time, which kinda threw me off to making wrong conclusions.
Thanks again and wish you nice holidays.