Low rate :error return when posting Sensor Reading

[ Posting in Software as there doesn’t appear to be a category for the my.farm.bot service infrastructure ]

Intermittently seeing this error when my Sequence tries to record a Sensor Reading :slight_smile:

@jsimmonds I’m not seeing anything in our logs to indicate that there’s a problem with the endpoint. Can you tell me a bit more about how that sequence works? I’ve seen this happen when users have daisy chained routers or bad NAT configuration, but usually the AMQP connection is the first one to exhibit the problem. I know that our hosting vendor (Heroku) will cut off all HTTP connections that last longer than 30 seconds, but it seems unlikely that a sensor reading would take that long :thinking:

Any additional information would be appreciated.

Thanks @RickCarlino. Sorry about the lack of detail; guess I just assumed that someone would be aware of this :wink:

First I’ll check that the Sensor Readings with those UUIDs were actually persisted on the Web App.
Next, I’ll try to get a tcpdump capture or other trace that helps us diagnose.

Must say I’m often surprised how well the global IP connection works every day between AU and US !

| edit |
These logs are very low rate ( <= 1% of total Sensor Readings )
My simple Sequence runs every 1 minute and takes 1 Sensor Reading ( analogue ) per run.
All readings get safely persisted, so it looks like an innocuous protocol or software bug.

| final edit |
My curiosity is waning ( life is too short and there are many higher priority things beyond this intriguing rabbit hole :slight_smile: ) but want to add a final bit of evidence.

Here’s a pic. of 1 Sequence execution where the trailing HTTP error occurs. It usually logs around 1…2 seconds after the Sensor Reading was actually acquired.

1 Like

I agree, “Remote connection closed” errors have historically been random network flukes. If we start getting more reports in the wild, I can dig deeper. My guess is it is just a transient failure due to the network.

Possibly . . did you mention Heroku runs your API service endpoints ? SDN is the networking paradigm-de-jour ! Some network support person probably needs to adjust a little “packet workflow” somewhere :wink:

@jsimmonds Correct. There are definitely some “Heroku-isms” that could be at play. For instance, in Heroku, you can’t make a request longer than 30 seconds, after that the connection is timed out.


I hate smearing (misrepresenting) languages . . du jour is what I meant to write :slight_smile:

1 Like