Cyclic Data Loss Issue with Codesys-MQTT, Telegraf, and InfluxDB

[JeromeH60]

Hello,

I am collecting several analog inputs on a Wago Edge Computer, which hosts a Docker environment with Mosquitto, Telegraf, InfluxDB, and Grafana. I have implemented the Codesys-MQTT library, and I can successfully record data in InfluxDB. However, I am missing a certain number of data points in a cyclic manner (I should have one data point every 30ms).

I have simplified my program to ensure that the issue does not come from my processing, and everything seems correct.

Could you tell me if you see any issues and whether I have proceeded correctly?

Thank you for your help

[JeromeH60]

My main issue is ok now, my other issue is that InfluxDB doesn’t seem to store values of "0", but I need them to be recorded. I need to look into how to make that happen.

[stefandreyer]

Hello,

okay, Interesting setup.which qos do you use for your pakages? how many data point do you process?

br stefan

[JeromeH60]

Hello,

I've set the QOS to “0” to be able to send all the points in time. I should be able to record about 50 measurements every 30ms.

[stefandreyer]

Hello,

Looks clear, QoS 1 and 2 needs more cycles for packet handling, sa you may loos data...

how is the cycle time of your plc?

[JeromeH60]

My PLC cycle is set to 15ms because I want to be able to record a point every 30ms (I do, however, have a processing function that allows me to limit the number of recorded data if I see that the value has changed very little).
I'm obliged to set the cycle time to 15ms because publication only works on the rising edge of the “send” input. Is it possible to publish without having to go through a rising edge of send?

[stefandreyer]

Hi,

beacause of tht MQTT is maybe not the right protokoll for real time data I implementet this with trigger. It's more to exchange states and this secure with QoS level >0. But If you want to achive a Publish every cycle you may call a second publisher with a negetiv puls to the other publisher on the same topic, so one cycle pubisher 1 sends next cycle publisher 2 sends, so one publish each cycle. In this way you be be able to use QoS > 0 with more publishers and some kind of publisher counter....

BR Stefan

[stefandreyer]

Hi, a other approche, wich I sometimes use for other FBs too, is to call the fb a second time, after send was set to falso, so that for the next cycle a new rising edge is possible...

br Stefan

[JeromeH60]

Hello, the system now seems to be working properly, I'll let it live for a few weeks.
Codesys currently publish around 150 data per second for recording on influxdb.
In any case thank you for your work on this CODESYS-MQTT library which allows me to realize my project.