Allop

There is one way this could happen, which is that the process that has the port open (ICC in this case) doesn't properly die. Restarting that process only starts a second instance, which then cannot communicate with the port because it is held open by the first process. Restarting the entire machine, of course, kills all the processes including the errant one holding the port open, and all is well again.
So some things you could do. First thing would be to run something like "top" to see if ICC is still running. Or list all the processes with the command "ps aux" and see if it is in the list. You could also check what has the serial port open with fuser, for example, on my machine I can see that screen (a terminal aggregator) is attached to the relevant serial port:
$ sudo fuser -v /dev/ttyUSB0 USER PID ACCESS COMMAND /dev/ttyUSB0: plonky 13848 F.... screen And then you can kill that process, using "kill 13848" in this example, and if it stubornly refuses to die, you can take out the big hammer and smack it with a kill signal: "kill -KILL 13848"
A discussion about how signals in unix work is beyond this post... suffice it to say that unix is a bit more polite... by default it asks processes to terminate rather than shooting it. On windows... using TerminateProcess is akin to always smacking things with the biggest hammer... that is to say, this is a feature, not a bug :-)

Cool quite a super piece of software:
 

I am not sure why it stops and that does make it difficult. In my case what I did was to stop the upload of data to emon.org the moment it detected that there is no connection to emon.org. 
I am sending a ping to the .org site every x-time (Cant remember the time) and if it times out - stop sending until the ping goes through again. It helps but I still get bad data on the .org site. I used to use the .org site for remote view of the system , that way I didnt need to port forward anything but now a days I use Openhab for that but it still an easy way to remote view.

I know that a couple of the gents installed a permanent fan on the inverter that has made quite a difference. That could be an easy test to do. The data jumps I noticed that some values will fall to 0 but if you look at the trend , I would venture that it is more of an data issue than a hardware issue. 
I found that the data sent to emon.org sometimes goes gaga. In my case Ill get some spikes from time to time , but only on the .org site. 

There is of course a way to fake the serial number of a pi and put the old one back in... but I'm not telling, and I only know by accident because I messed around in the guts of the system :-)

Just a short note to thank Manie for constantly striving to make ICC better and more stable.
I think from the screenshot attached (Early Saturday morning) we can all see that ICC Pi is running stable with no issues.

Keep up the good work.

Running ICC Pi V2.2.1;
Axpert 3kVa; Local emon; EmonWeb; PVOutput.org and MQTT  

I think it happens when guys get frustrated with clouds moving in front of the sun and bangs the inverter with the head. 

Jissie Manie, I think you're the one who need to get a thank you. The "testing" I had to do was to install the Windows ICC and run it. No problems whatsoever so far and yet you keep updating with new features......Thank YOU!

Sent from my S60 using Tapatalk

REXX?  ... is a scripting language available for many OSs, also for Linux and Raspian. I tried to modify the autostart script that come with aicc.  Yes, i tried!  with no success. So i searched and installed REXX.
This is now my ICC start script:
#!/usr/bin/rexx trace 'o' /* started by autorun: in pi/.config/lxsession/LSDE-pi/autostart */ dir = '/home/pi/AICC' log = dir'/RunAICC.log' start_time = time('E') /* start stop watch */ logt = 'ICC started at' date() time() call lineout log, logt /* write to log */ say logt rc=chdir(dir) /*go to aicc directory and start aicc*/ address cmd './ICC' /* start app */ /* calc run duration hrs min */ duration = time(R) /* read stop watch */ hours = duration%3600 /* get hrs only */ minutes = (duration//3600) % 60 /* from rest calc mins */ logt = 'ICC ended at' date() time() '- duration:' hours'hrs' minutes'min' say logt /* say to console */ call lineout log, logt call lineout log, '- - - - - - -' Log file sample:
ICC started at 14 Feb 2017 18:42:01 ICC ended at 14 Feb 2017 18:42:01 - duration: 15hrs 51min Isn't it nice and easy to read?  Rexx is typeless, all conversions are done at the time its needed - Automatically.  
---------------- 
So far - so good.  I tried to autostart this script by adding it to pi/.config/lxsession/LSDE-pi/autostart.  That did not work. Is there no tool/gui that helps to configure autostart of programs?  There also must be a list of programs that already start automatically.. 

Hi.
Let's attach here the results of ICC.
Here are mine!
http://www.pvoutput.org/list.jsp?id=55186&sid=50118
and
https://emoncms.org/dashboard/view?id=39626
Thank you for the great software!

Hi.
Let's attach here the results of ICC.
Here are mine!
http://www.pvoutput.org/list.jsp?id=55186&sid=50118
and
https://emoncms.org/dashboard/view?id=39626
Thank you for the great software!