The two patches to fix ethernet stability have been merger into the development branch, and also some changes to the gateway logging. I updated the build instructions to reflect the changes.

@masmat This is my mosquitto.log

1521680648: mosquitto version 1.4.10 (build date Fri, 22 Dec 2017 08:19:25 +0000) starting
1521680648: Config loaded from /etc/mosquitto/mosquitto.conf.
1521680648: Opening ipv4 listen socket on port 1883.
1521680648: Opening ipv6 listen socket on port 1883.
1521762617: Saving in-memory database to /var/lib/mosquitto/mosquitto.db.
1521763549: New connection from 127.0.0.1 on port 1883.
1521763549: New client connected from 127.0.0.1 as mygateway1 (c1, k15, u'rpi').

A socket error message:

1521764339: Socket error on client mygateway1, disconnecting.

shows only if I stop/exit mysgw.

@gohan Sorry to hear that, mysgw doesn't change any system files, a simple reboot and rebuild of the gateway should be enough to resume communication in a case of more serious error.

@mvader The gateway will respond with an I_VERSION every time a controller connected to port 5003 sends an I_VERSION request.

@masmat I did a test with the following configuration:

• Raspberry Pi 1 with Raspbian (2017-11-29-raspbian-stretch-lite)

  pi@raspberrypi:~/MySensors $ lsb_release -a
  No LSB modules are available.
  Distributor ID: Raspbian
  Description:    Raspbian GNU/Linux 9.1 (stretch)
  Release:        9.1
  Codename:       stretch
  

• MySensors master branch - Protocol version - 2.2.0

  ./configure --my-transport=nrf24 --my-gateway=mqtt --my-controller-ip-address=127.0.0.1 --my-mqtt-publish-topic-prefix=mysensors-out --my-mqtt-subscribe-topic-prefix=mysensors-in --my-mqtt-client-id=m ygateway1 --my-mqtt-user=rpi --my-mqtt-password=password
  

• Mosquitto version 1.4.10 (build date Fri, 22 Dec 2017 08:19:25 +0000) [installed from apt-get]

  pi@raspberrypi:~/MySensors $ cat /etc/mosquitto/mosquitto.conf
  # Place your local configuration in /etc/mosquitto/conf.d/
  #
  # A full description of the configuration file is at
  # /usr/share/doc/mosquitto/examples/mosquitto.conf.example
  
  pid_file /var/run/mosquitto.pid
  
  persistence true
  persistence_location /var/lib/mosquitto/
  
  log_dest file /var/log/mosquitto/mosquitto.log
  
  allow_anonymous false
  password_file /etc/mosquitto/passwordfile
  
  include_dir /etc/mosquitto/conf.d
  

  Added the mosquitto user (user=rpi, password=password) with:

  sudo mosquitto_passwd -c /etc/mosquitto/passwordfile rpi
  

It didn't show any error to connect to the MQTT broker:

pi@raspberrypi:~/MySensors $ sudo ./bin/mysgw -d
mysgw: Starting gateway...
mysgw: Protocol version - 2.2.0
mysgw: MCO:BGN:INIT GW,CP=RNNGL---,VER=2.2.0
mysgw: TSF:LRT:OK
mysgw: TSM:INIT
mysgw: TSF:WUR:MS=0
mysgw: TSM:INIT:TSP OK
mysgw: TSM:INIT:GW MODE
mysgw: TSM:READY:ID=0,PAR=0,DIS=0
mysgw: MCO:REG:NOT NEEDED
mysgw: MCO:BGN:STP
mysgw: MCO:BGN:INIT OK,TSP=1
mysgw: GWT:RMQ:MQTT RECONNECT
mysgw: connected to 127.0.0.1
mysgw: GWT:RMQ:MQTT CONNECTED
mysgw: GWT:TPS:TOPIC=mysensors-out/0/255/0/0/18,MSG SENT
mysgw: TSF:MSG:READ,5-5-0,s=255,c=3,t=24,pt=1,l=1,sg=0:1
mysgw: TSF:MSG:PINGED,ID=5,HP=1
mysgw: TSF:MSG:SEND,0-0-5-5,s=255,c=3,t=25,pt=1,l=1,sg=0,ft=0,st=OK:1
mysgw: TSF:MSG:READ,5-5-0,s=255,c=3,t=32,pt=5,l=4,sg=0:500
mysgw: GWT:TPS:TOPIC=mysensors-out/5/255/3/0/32,MSG SENT
mysgw: TSF:MSG:READ,5-5-0,s=255,c=3,t=33,pt=5,l=4,sg=0:4951
mysgw: GWT:TPS:TOPIC=mysensors-out/5/255/3/0/33,MSG SENT
mysgw: TSF:MSG:READ,5-5-0,s=1,c=1,t=37,pt=4,l=4,sg=0:36053
mysgw: GWT:TPS:TOPIC=mysensors-out/5/1/1/0/37,MSG SENT
mysgw: TSF:MSG:READ,5-5-0,s=255,c=3,t=32,pt=5,l=4,sg=0:500
mysgw: GWT:TPS:TOPIC=mysensors-out/5/255/3/0/32,MSG SENT
mysgw: TSF:MSG:READ,5-5-0,s=255,c=3,t=33,pt=5,l=4,sg=0:5000
mysgw: GWT:TPS:TOPIC=mysensors-out/5/255/3/0/33,MSG SENT
mysgw: TSF:MSG:READ,5-5-0,s=1,c=1,t=37,pt=4,l=4,sg=0:36055
mysgw: GWT:TPS:TOPIC=mysensors-out/5/1/1/0/37,MSG SENT
mysgw: TSF:MSG:READ,5-5-0,s=255,c=3,t=32,pt=5,l=4,sg=0:500
mysgw: GWT:TPS:TOPIC=mysensors-out/5/255/3/0/32,MSG SENT
mysgw: TSF:MSG:READ,5-5-0,s=255,c=3,t=33,pt=5,l=4,sg=0:5000
mysgw: GWT:TPS:TOPIC=mysensors-out/5/255/3/0/33,MSG SENT
mysgw: TSF:MSG:READ,5-5-0,s=1,c=1,t=37,pt=4,l=4,sg=0:36055
mysgw: GWT:TPS:TOPIC=mysensors-out/5/1/1/0/37,MSG SENT

@gieljnssns One way is to replace the changed files, in this case EthernetServer.cpp for the first patch and EthernetClient.cpp for the second patch.

You can download from:
https://raw.githubusercontent.com/marceloaqno/MySensors/linux-fix-stability/drivers/Linux/EthernetServer.cpp
https://raw.githubusercontent.com/marceloaqno/MySensors/linux-fix-stability/drivers/Linux/EthernetClient.cpp

Here is a step-by-step:

git clone https://github.com/mysensors/MySensors.git
cd MySensors
curl https://raw.githubusercontent.com/marceloaqno/MySensors/linux-fix-stability/drivers/Linux/EthernetServer.cpp -o drivers/Linux/EthernetServer.cpp
curl https://raw.githubusercontent.com/marceloaqno/MySensors/linux-fix-stability/drivers/Linux/EthernetClient.cpp -o drivers/Linux/EthernetClient.cpp

from here just run ./configure with your options and then make

@masmat What is the mqtt broker / version you are using?

@otto001 The second patch can be applied without the first one.

@gieljnssns The fixes I mentioned above haven't yet been applied to the development branch. Eventually they will be, but I'm still doing some testing.

@otto001 After a few hours running the reconnect script, I noticed several error messages in my gateway log:

accept(): Too many open files

The problem was that the gateway wasn't releasing the socket descriptor after each disconnection of the controller. After reaching the limit (mine was 1036) the controller can't connect.

I'm not entirely sure that this problem is related to what you're having. You can check how many file descriptors currently opened by the gateway with the command below

sudo lsof -u root |grep mysgw |wc -l

The fix:
https://github.com/marceloaqno/MySensors/commit/a40d4441b7460225100398ff6f2581c2b0df36ea

0 marceloaqno committed to marceloaqno/MySensors

Linux: Fix stability problem with ethernet (2)

Fix bug with tcp sockets where socket descriptor wasn't being freed.