💬 Building a Raspberry Pi Gateway

[otto001]

@marceloaqno : of course, thanks for the hint!

I had the following configure:

./configure --my-transport=nrf24 --my-rf24-irq-pin=15 --my-gateway=ethernet --my-port=5004

(The high CPU problem also occured before I connected PIN 15 to IRQ without the corresponding config option. I hoped to solve the Problem using IRQ, but this did not help. And I have a second GW with default port 5003 in my network. This is running an ancient version of Raspi GW Software and this is working fine -but slower). I am using current fhem as controller and I have it connecting every 30 minutes to the gateways due to stability issues in the past. There are only about 6 nodes (just motion, humidity & temp and lux) connected to this gateway.

In addition to that I changed the default channel in MyConfig.h (because I have two different MySensor networks):

#ifndef MY_RF24_CHANNEL
#define MY_RF24_CHANNEL (82)
#endif

I am pretty sure, that this is not important, but I also set up rsyslog to write gw logs to an own file instead of syslog....

Maybe this helps.

Cheers,
Otto

[marceloaqno]

@otto001 I think I found the problem. I wrote a fix for it:
https://github.com/marceloaqno/MySensors/commit/ca8865d8e1fe1b08195943fae1640d5fddb81924

[otto001]

@marceloaqno : Wow, you are fast! Thanks!
Just applied the patch and recompiled. I will report in a couple of days while my monitoring-script is still running (and logging)...

Cheers,
Otto

[mvader]

@marceloaqno I am also recompiling and testing.
thanks for the (hopeful) fix
I will report back with my status in a few days

[otto001]

@marceloaqno :

At first, THANKS for your help. CPU usage does not rise anymore.
But unfortunately no more node READs after the same situation:

Mar  4 07:00:09 raspiez mysgw: TSF:MSG:READ,14-14-0,s=255,c=3,t=6,pt=1,l=1,sg=0:0
Mar  4 07:00:09 raspiez mysgw: TSF:MSG:READ,12-12-0,s=255,c=3,t=6,pt=1,l=1,sg=0:0
Mar  4 07:00:09 raspiez mysgw: GWT:RFC:C=1,MSG=14;255;3;0;6;M
Mar  4 07:00:09 raspiez mysgw: !TSF:MSG:SEND,0-0-14-14,s=255,c=3,t=6,pt=0,l=1,sg=0,ft=0,st=NACK:M
Mar  4 07:00:09 raspiez mysgw: TSF:MSG:READ,11-11-0,s=255,c=0,t=18,pt=0,l=3,sg=0:1.5
Mar  4 07:00:09 raspiez mysgw: GWT:RFC:C=1,MSG=12;255;3;0;6;M
Mar  4 07:00:09 raspiez mysgw: TSF:MSG:SEND,0-0-12-12,s=255,c=3,t=6,pt=0,l=1,sg=0,ft=0,st=OK:M
Mar  4 07:00:13 raspiez mysgw: TSM:READY:NWD REQ
Mar  4 07:00:13 raspiez mysgw: TSF:MSG:SEND,0-0-255-255,s=255,c=3,t=20,pt=0,l=0,sg=0,ft=0,st=OK:
Mar  4 07:00:13 raspiez mysgw: TSF:SRT:OK
Mar  4 07:00:13 raspiez mysgw: TSF:SAN:OK
Mar  4 07:00:37 raspiez mysgw: Ethernet client disconnected.
Mar  4 07:00:37 raspiez mysgw: New connection from 192.168.1.11
Mar  4 07:00:37 raspiez mysgw: GWT:TSA:C=0,CONNECTED
Mar  4 07:00:37 raspiez mysgw: GWT:TSA:C=1,DISCONNECTED

After this point (again after a reconnect from the controller) no more messages from any node PLUS the cpu usage stayed normal (what makes it even worse at this moment, as I would have to create a script which monitors the logfile and restarts the mysgw service in this situation). As soon as the service is restarted, everything is working fine again.

What happens after this point in the log:

Mar  4 07:00:37 raspiez mysgw: GWT:RFC:C=0,MSG=0;0;3;0;2;
Mar  4 07:15:13 raspiez mysgw: TSF:SAN:OK
Mar  4 07:20:13 raspiez mysgw: TSM:READY:NWD REQ
Mar  4 07:20:13 raspiez mysgw: TSF:MSG:SEND,0-0-255-255,s=255,c=3,t=20,pt=0,l=0,sg=0,ft=0,st=OK:
Mar  4 07:30:13 raspiez mysgw: TSF:SRT:OK
Mar  4 07:30:13 raspiez mysgw: TSF:SAN:OK
Mar  4 07:30:37 raspiez mysgw: GWT:TSA:C=0,DISCONNECTED
Mar  4 07:30:37 raspiez mysgw: Ethernet client disconnected.
Mar  4 07:30:37 raspiez mysgw: New connection from 192.168.1.11
Mar  4 07:30:37 raspiez mysgw: GWT:TSA:C=0,CONNECTED
Mar  4 07:30:37 raspiez mysgw: GWT:RFC:C=0,MSG=0;0;3;0;2;
Mar  4 07:40:13 raspiez mysgw: TSM:READY:NWD REQ
Mar  4 07:40:13 raspiez mysgw: TSF:MSG:SEND,0-0-255-255,s=255,c=3,t=20,pt=0,l=0,sg=0,ft=0,st=OK:
Mar  4 07:45:13 raspiez mysgw: TSF:SAN:OK
Mar  4 08:00:13 raspiez mysgw: TSM:READY:NWD REQ
Mar  4 08:00:13 raspiez mysgw: TSF:MSG:SEND,0-0-255-255,s=255,c=3,t=20,pt=0,l=0,sg=0,ft=0,st=OK:
Mar  4 08:00:13 raspiez mysgw: TSF:SRT:OK
Mar  4 08:00:13 raspiez mysgw: TSF:SAN:OK
Mar  4 08:00:37 raspiez mysgw: GWT:TSA:C=0,DISCONNECTED
Mar  4 08:00:37 raspiez mysgw: Ethernet client disconnected.
Mar  4 08:00:37 raspiez mysgw: New connection from 192.168.1.11
Mar  4 08:00:37 raspiez mysgw: GWT:TSA:C=0,CONNECTED

and so on....

It seems that the increasing cpu usage is only a symptom for the real problem.
As I am using this in production I will dis-apply the patch and use the daemon as before (beacuse using the python script for cpu-usage-monitoring makes the situation somehow controllable for me).
Maybe you could have an additional eye on this please? Could I help you with further information? Maybe I could add some additional logging somewhere in the code?

Thanks again and cheers,
Otto

[marceloaqno]

@otto001 I'm trying to reproduce this problem, but so far I've got nothing beyond the normal behavior of mysgw.
I tried with a RPi1 and RPi3 using the fhem controller with the latest raspbian.

[mvader]

@marceloaqno said in 💬 Building a Raspberry Pi Gateway:

@otto001 I'm trying to reproduce this problem, but so far I've got nothing beyond the normal behavior of mysgw.
I tried with a RPi1 and RPi3 using the fhem controller with the latest raspbian.

My uptime is at 4 days now. so far still working with your patch. CPU is at 2% and sensor network still working correctly. will report if that changes.

[Nick Willis]

I've been using your restarting method and its been working ok and has done one restart in 5 days.
However I now see something else where altho cpu use is fine, none of my nodes can communicate to the gateway. My controller can connect to the gateway no problem but there are no messages received. Restarting nodes does not help. Only a manual restart of the gateway gets things going again - at least for a while anyway.
Where is the log output from the gateway to be found? I know it can be run from the bash prompt to write the output to screen but how about logging the installed service output?

[mfalkvidd]

@nick-willis the log is sent to syslog by default. See https://www.mysensors.org/build/raspberry#troubleshooting for details.

[otto001]

@marceloaqno : This is strange. I am using a RPi 1 on this gateway with raspbian Jessie. Beside the GW-service there is only a small python programm which listens on an UPD-Port and plays an mp3 if it receives suitable data (for doorbell). Since March 4, 20:15 there were several restarts due to high cpu usage (after dis-applying your patch):

2018-03-04 23:01:03 cpu (99.7) very high! --> restarting service
2018-03-05 07:31:03 cpu (99.7) very high! --> restarting service
2018-03-05 11:01:03 cpu (98.7) very high! --> restarting service
2018-03-05 13:31:04 cpu (94.7) very high! --> restarting service
2018-03-06 11:31:03 cpu (99.7) very high! --> restarting service
2018-03-06 12:31:03 cpu (98.7) very high! --> restarting service
2018-03-07 06:01:03 cpu (94.8) very high! --> restarting service
2018-03-07 22:31:03 cpu (99.8) very high! --> restarting service

With this "auto-restarts" the sensors are working fine, I did not notice any problems.

However, I just upgraded to the recent Stretch (what takes a LONG time on RPi 1 and made me struggle a little bit with the new network names). I also rebuilt the GW of course and will report here what will happen....

@Nick-Willis : did you apply the patch of marceloaqno or not? Did you enable debugging? (see here )

Cheers,
Otto

[marceloaqno]

@otto001 In a new attempt to trigger this problem, I am using a script that will force the fhem controller to reconnect every 30s to the gateway (built with mysensors master branch, unpatched), and sends me a note in case of high CPU usage:

I'll leave it running for a few days to see what happens.

Here is the script, if anyone is interested:

#!/bin/sh

while true; do
  (echo "set mysgw disconnect"; echo "quit") | telnet localhost 7072 > /dev/null 2>&1
  sleep 1
  (echo "set mysgw connect"; echo "quit") | telnet localhost 7072 > /dev/null 2>&1
  sleep 1

  cpu_percent=$(ps -C "mysgw" -o %cpu=)
  cpu_percent=${cpu_percent%%.*}
  if [ "$cpu_percent" -ge 90 ]
  then
    echo "ALERT: HIGH CPU USAGE"
    # pushbullet script from https://gist.github.com/outadoc/189bd3ccbf5d6e0f39e4
    /home/pi/pushbullet.sh "RPi1: HIGH CPU USAGE"
    exit
  fi

  sleep 30
done

[otto001]

@marceloaqno : Thank you!
Just a notice: I do NOT disconnect, I just connect every 30 minutes. Has been working for me since more than 2 years (ancient ms-gw version on RPi).

Yesterday I did a dist-upgrade to stretch on the "problematic" gw (as mentioned above) - without your patch but with my monitoring-script running. It is a little bit to early I presume, but no high cpu problems so far. I will report back here in a few days. Maybe this problem is really related to old libraries or something?! If this is running stable for a few days I will apply your patch and see what happens.
Thanks again!

Cheers,
Pula

[marceloaqno]

@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

[A Former User]

How do I update from master branch to development branch?

[gohan]

Why do you want to do that? Development version is still work in progress

[A Former User]

I'm using a raspberry pi gateway in home-assistant.
Sometimes my gateway doesn't work anymore, and when I restart the service, it is running good
I used ./configure --my-transport=nrf24 --my-rf24-irq-pin=15 --my-gateway=ethernet --my-port=5003

[gohan]

Read the article again, there is the instruction on how to download the development version, but before you have to uninstall the mysgw service and rename the current mysensors folder to keep as backup.

[marceloaqno]

@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]

@marceloaqno :
Thank you!!! Should this patch be applied together with your first patch?
Interesting:
After some days without problems on the newest raspbian packages the gw hung again, but WITHOUT high cpu usage (and WITHOUT your first patch applied). There were just no more reads from the sensors in the log, only the re-connects of fhem. After restarting the service, everything is fine again.
Unfortunately I did see your post after that :-(
But I checked the log and could not find the words "too" or "Too" in the log (running service with -d).
Some minutes after restarting I checked how many open files there were with the command you supplied: 21

In the meantime I have created a small udp listening service which restarts the service if a certain string is received, so I can restart the service by double-clicking one of my switches without the need of login to the gw (also for the wife)....

@gieljnssns : I think you are experiencing exactly the same problem we are discussing :-)

Thanks again!
Cheers,
Otto

[gohan]

I'm using dietpi and I am not having big issues even running 2 gateway instances with 2 different radios