Refused connects

[zoneblue]

Bad news Bob, the refused modubus over ethernet connects are still present here using 1609.

I havent tried quickening the sample rate from the current 2 minute interval, but regardless i still get about the same number of refused connections that i got with 1370.

At present thats about 1 every hour. And they are still of the 'refused' type as opposed to 'ignored'.

And the ethernet lockups still come less frequently about once a 10 days or so.

Changing from 1 min interval to 2 minutes made a big difference, but didnt completely eradicate either symptom.

Back to the drawing board... Is there anything i can do to help diagnose this further?

Im 99% sure we have ruled out newmodbus, as the issue was worse using the python lib. It does seem though to be a combination of ARM resources and classic firmware.

The one thing i have left to try is my old 400Mhz Eden 586 board. Will try to find it soon.

Tom and Ross, have you anything new to report on this one?

[TomW]

Tom and Ross, have you anything new to report on this one?

My last reset was 10-8 and it is 10-29 today so 3 weeks with no resets.  It was my solar Classic that was doing it and it is also running the latest firmware for the WB jr mounted on its shunt now.

That is what has been going on here.

Maybe got past it swapping out a D-link router that had other issues for a Linksys WRT54G/L right after that last reset?

Tom

[RossW]

Im 99% sure we have ruled out newmodbus, as the issue was worse using the python lib. It does seem though to be a combination of ARM resources and classic firmware.

The one thing i have left to try is my old 400Mhz Eden 586 board. Will try to find it soon.

Tom and Ross, have you anything new to report on this one?

I can report zero failures in months and months, but I'm running FreeBSD with a gruntier processor:

  CPU: VIA Esther processor 1200MHz (1200.01-MHz 686-class CPU)
     Origin = "CentaurHauls"  Id = 0x6a9  Family = 6  Model = a  Stepping = 9

[TomW]

I can report zero failures in months and months, but I'm running FreeBSD with a gruntier processor:

  CPU: VIA Esther processor 1200MHz (1200.01-MHz 686-class CPU)
     Origin = "CentaurHauls"  Id = 0x6a9  Family = 6  Model = a  Stepping = 9

Bragging or complaining? 

Tom

[RossW]

Bragging or complaining? 

Yes   

[zoneblue]

Persistant that i am, im still working on this. Weve always presumed that the problem mightnt occur if the modbus connection was left open permanently, i can now confirm that it seems to work just fine on theh cubieboard.

Code Select Expand


./newmodbus-1.0.19-ARM -i60 -l/home/blackbox/test/ 192.168.0.223  4117 4371  >> /home/blackbox/test/log.txt &


Load average around 0.05, logging at 1second intervals. Been running for a couple of days without any observable problems. Graph below of classic output current, and output current less WbJr current, ie load current (more or less). You can see the WBjr data is more granular.

Also ive asked stephen for more info about his 'keep alive' workaround mentioned here:
http://midniteforum.com/index.php?topic=738.msg4920#msg4920

[zoneblue]

Heres a further interesting thing. In my discussions with the various people that have encountered this modbus connect issue, one really strong common denominator to avoiding the symptom is longer sample times.

So i thought is was high time i actually tried this. So for the last 24 hours cubieboard has been running at 5 minute intervals. And the result is: not one refused connect, a 100% complete 24 hour dataset. To be clear i have never _ever_ seen this before on that board.

So something about timing is clearly a factor here. Is the classic better able to clean up closed conections, given enough time? I guess ill reduce it by one minute a day to see where the cut off point lies.

The problem is that 5 minute samples create a very blurry picture of throughput on the controller. The fridge and waterpump square wave forms look like sine waves on 5min samples, and the resulting inaccuracy isnt going to help things like SOC.

I want to make clear that i think the future of blackbox depends on resolution of this issue.

However i have asked Ross to consider some changes to newmodbus to better utelise its single open connection mode, and that is one route we can certainly take.

However in my corespondance with people i see a pattern of people each quietly doing various hacks to make their own setup hobble along, then forget about the issue, leaving it for the next person to stumble over. This is something that doesnt sit well with me, and i would very much prefer to work with you if i can, to help fix it properly.