[KNX] Too much tunnel connections

[teaserrr]

I regularly notice that I cannot make a tunnel connection in ETS to the KNX bus anymore. I can resolve this by starting OpenHAB. After restarting, I have no problems for several weeks.

Today I had this issue again and I listed all the UDP connections:

$ sudo netstat -aunp | grep java
udp6       0      0 :::5353                 :::*                                682/java
udp6       0      0 192.168.0.180:37610     :::*                                682/java
udp6       0      0 192.168.0.180:38186     :::*                                682/java
udp6       0      0 192.168.0.180:36204     :::*                                682/java
udp6       0      0 :::1900                 :::*                                682/java
udp6       0      0 192.168.0.180:49556     :::*                                682/java
udp6       0      0 192.168.0.180:52725     :::*                                682/java

Then I restarted OpenHAB. Running the above command gave no results.
During startup, connections were created again:

$ sudo netstat -aunp | grep java
udp6       0      0 192.168.0.180:51413     :::*                                25775/java
udp6       0      0 :::5353                 :::*                                25775/java
udp6       0      0 :::1900                 :::*                                25775/java

After the KNX binding was initialized, and I started seeing loads of status telegrams in ETS group monitor (which always happens during startup), 1 more connection was created:
$ sudo netstat -aunp | grep java

udp6       0      0 192.168.0.180:51413     :::*                                25775/java
udp6       0      0 :::5353                 :::*                                25775/java
udp6       0      0 :::1900                 :::*                                25775/java
udp6       0      0 192.168.0.180:33478     :::*                                25775/java

So that's 4 UDP connections vs 7 before restart. We can assume at this point there is only 1 for KNX. My KNX IP interface has 4 tunnels, so 3 extra tunnels would mean all tunnels are occupied and any extra (e.g. ETS) cannot be established.

In the OpenHAB knx logging I sporadically see disconnects and reconnects. However sometimes it tries to reconnect multiple times in the same 100ms:

2019-11-06 10:02:00.549 [WARN ] [nx.internal.connection.KNXConnection] - KNX link has been lost (reason: server request on object tunneling link link (closed) 192.168.0.187:3671 TP1 medium, device 0.0.0, hopcount 6)
2019-11-06 10:02:00.553 [INFO ] [nx.internal.connection.KNXConnection] - KNX link will be retried in 30 seconds
2019-11-06 10:02:00.598 [WARN ] [nx.internal.connection.KNXConnection] - KNX link has been lost (reason: server request on object tunneling link link (closed) 192.168.0.187:3671 TP1 medium, device 0.0.0, hopcount 6)
2019-11-06 10:02:00.603 [INFO ] [nx.internal.connection.KNXConnection] - KNX link will be retried in 30 seconds 
...
2019-11-06 10:02:30.562 [INFO ] [nx.internal.connection.KNXConnection] - Trying to (re-)connect to KNX...
2019-11-06 10:02:30.615 [INFO ] [nx.internal.connection.KNXConnection] - Established connection to KNX bus on 192.168.0.187:3671 in mode TUNNEL.
2019-11-06 10:02:30.608 [INFO ] [nx.internal.connection.KNXConnection] - Trying to (re-)connect to KNX...
2019-11-06 10:02:30.646 [INFO ] [nx.internal.connection.KNXConnection] - Connected to KNX
2019-11-06 10:02:30.680 [INFO ] [nx.internal.connection.KNXConnection] - Established connection to KNX bus on 192.168.0.187:3671 in mode TUNNEL.
2019-11-06 10:02:30.697 [INFO ] [nx.internal.connection.KNXConnection] - Connected to KNX 

Could this be the cause? Or are the old tunnels somehow not released?

Your Environment

OpenHab 2.4.0 Release
binding-knx1 - 1.13.0
Running on Raspberry Pi 3
KNX IP interface: MDT SCN-IP100.02

knx.conf:

ip=192.168.0.187
ignorelocalevents=true
type=TUNNEL
localIp=192.168.0.180
autoReconnectPeriod=30

[quensen]

Did you try downgrading to org.openhab.binding.knx-1.11.0.jar?
That solved it for me so far...

[teaserrr]

Thanks for the suggestion. I've just downgraded the binding so I'll see how it works out.

[quensen]

Hi @teaserrr,

I need to withdraw my suggestion. I tried several KNX Binding releases 1.9 to 1.14 and after same runtime, I get these errors again. But these errors appeared weeks ago. They weren't there a few months ago. Strange...

Maybe, it's something server-related?
I am running OH 1.8.2 on a QNAP NAS with the most current firmware. You?

Regards,
Stefano

[teaserrr]

@SAQ1973 I'm running on an RPI 3B with Openhabian.

I havan't tried anymore since I downgraded and at the moment I can't check in ETS because I'm not on my home network and I can't access the bus over my VPN. But I can see that there are still only 4 UDP 'connections' when I log in to my RPI:

udp6       0      0 192.168.0.180:51413     :::*                                25775/java
udp6       0      0 :::5353                 :::*                                25775/java
udp6       0      0 :::1900                 :::*                                25775/java
udp6       0      0 192.168.0.180:43929     :::*                                25775/java

I also see 2 disconnects/reconnects since last time I restarted OH. In these cases there are no multiple reconnects in the same second as I originally reported. So for mee it seems solved by downgrading.

[quensen]

I noticed that my errors happened mostly (or maybe always) during the night. I will continue to monitor them.
I also reinstalled OH yesterday but keeping the configuration.
@teaserrr: Why is it udp6?