CONTROLLER RESET AND NO MYMIDNITE NO DATA

[GHOSTNH]

some time last night my lite got locked up @ 8pm  had to do a hard reboot this A.M. just to get the local app to find it .

reset all the parameters that were not right date ,time , amps limit, web enable ,gateway etc   now my midnite is not getting a data stream been several hrs ... what did i miss?? and any idea why the controller hung?

[atop8918]

Please take a look at the advanced toubleshooting area of mymidnite if you want to double check all the DNS settings:

http://mymidnite.com/?q=node/17

The communications dropping is a known bug. We've been working on it for a while now. Hopefully we'll have a fix soon.

[zoneblue]

Hi Andrew, have you been able to reproduce it in the lab? Im confident that if you could, it would be down hill from there.

The sure fire way would be to take one ARM developer board, install debian on it, and set up a cronjob every minute than opens, reads the entire register range, then closes. Log the results for 48 hours.  I havent done this for a while, but could check it again to make sure the latest firmware still does it, if that helps.

I dont know if its a bug, or just an incompatibility with certain linux stacks. The fact that Ryans mates NIGNN windows app reportedly works with out issue, given that it also works on open,close basis, suggests something.

But then almost everyone whose tryed ARM boards with linux has encoutnered it at some point. Sample rate was the key determinant, sample rates less than 3 mintutes triggered it. Which makes me wonder if it has something to do with not closing the connection properly, in that if you wait long enough the connection times out and allows you to start a new one. OTOH try again too soon and some buffer overflows or something.

Just thinking out loud. Really not trying to hassle you, just one coder to another.

[Halfcrazy]

I dont know if its a bug, or just an incompatibility with certain linux stacks. The fact that Ryans mates NIGNN windows app reportedly works with out issue, given that it also works on open,close basis, suggests something.

No Karl had all sorts of heartburn with the Classic as well. That was one of the biggest reasons he did the Connect, Read and disconnect. If he tried to keep it open the Classics Modbus would time out BUT never tell he TCP/IP stack to close. Thus leaving the port open so next time he talked to it he would get an acknowledgment from the network stack no modbus data. Very weird and caused lots of re doing code etc. He finally gave up and just closed the connection EVERY time to prevent the open port bug

This bug is still present today as best as I know. Andrew was looking into it last week but I am not sure if he found anything yet or not?

Ryan

[atop8918]

Hi, Zoneblue,
No worries! Please don't let my rants against certain other board members put you off -- I really am normally very reasonable and I'm very open to criticism and suggestions. There is a lot to criticize! I am an employee of MidNite Solar, though, and they are my bosses so I do what they tell me to do, not what certain folks on the forum tell me to do!

I have been able to reproduce some hangs albeit not reliably. I have found one large issue in that the advertise function was sometimes using up all the available connections on the Classic which I have fixed locally. I have high hopes that this is the notorious lockup issue but I don't have the empirical data to back that up yet.
I have not been able to reproduce the timeout-without-notification bug as all traces I've done at different rates show the correct TCP/IP behavior -- that is no lost FIN/ACK sequences or related issues. I've done this with a small C library I wrote on an rPI as well as with a 3rd party java library (jamod) on a PC. Now, that's not to say that these issues don't exist, they certainly do, it's just that I've had trouble reproducing them. It may simply be because I am not running the Classic in its full capacity-- I am just running the network stack itself on the Classic. There may be some subtler timing issues that need to be addressed. I am moving in that direction now so hopefully I'll get some traction off that!
So I'm working hard to get this (these) problems nailed so I can at least get out a build for Beta.
Thanks!
-Andrew

[GHOSTNH]

ANDREW
I followed all of the suggestion in the link about changeing the overrdie dns to 98.129.247.45 (was  at 0.0.0.0) and rebooting the system ...NO LUCK.. have to reset  date time amps limit and gateway every time i reset the unit .. still no data stream

[GHOSTNH]

forgot to add also rebooted all device in the network again

[GHOSTNH]

forgot to add can not do "custom mode" here.. ip are assiagned..  hard to believe that every time there is a issue one has to be in "custom mode" ..

[atop8918]

Hi, Ghost,
Sorry I don't understand -- why can't you do custom mode?

[GHOSTNH]

seems every few months this happens where something hangs and i have to reset everything and then i still get no data stream .. not very reliable .. for sure ..I have not even owned this controller for a year and have had nothing but hangs and disconnects and S/W issues and 1 bad bd from the beginning .. although im happy with the hardware performance.. the software  well lets just say seems it has lots of issues from what i see and read in these forums..

[Halfcrazy]

Custom mode defeats the dip switches and allows you to make changes with the local app. If it is not in custom mode then all the changes you make with the local app will be erased when the classic reboots.

This sounds like what you are experiencing. I am not sure why you "Can not" do custom mode? If the Local App is connected to it and you can see it then switching the switches to custom on the MNLP at that time with the classic running should not change anything IP wise and then you can write to it

[Halfcrazy]

I think the issue is a misunderstanding of how the Lite works. Unlike the Standard classic the dip switches write there info to the classic every time it is powered up. You will need the dip switches set to custom in order to use this classic like you are trying to use it.

[zoneblue]

Yes, on the lite you WANT to be in custom mode. However maybe you meant static ip?

Static IP for the classic is best, but not essential. To use static IP there are two ways:
-on your router, tell it to always assign the same ip address to the classic by its mac address
-on your router choose an ip address outside of the DHCP range. Most routers will issue addresses say from xxx.xxx.xxx.100 to xxx.xxx.xxx.150, hence an address xxx.xxx.xxx.90 is probably a good choice for the classic.

[boB]

I have found one large issue in that the advertise function was sometimes using up all the available connections on the Classic which I have fixed locally. I have high hopes that this is the notorious lockup issue but I don't have the empirical data to back that up yet.

Advertise can be disabled.  Maybe it would be a good idea to try this ?

EnableFlags3  register ADDRESS 4181,  Bit  0x0080

Advertise is Normally enabled if this bit is 0

Just set that bit, save EEprom by modbus or by going to any MNGP screen that saves Classic data,
and EnableFlags3  will also be saved to EEprom.

boB

[atop8918]

If you are experiencing these issues on a regular basis, Ghost, you may want to check your configuration. If you are setting anything other than the controls that the Lite provides through its DIP switches then you must be set to Custom mode. It seems you are setting current limits so you must set the Lite to Custom mode.
You should set a static IP address (using the Local App), set the Gateway and DNS settings and then reboot the unit. You may also want to enable the A-RST feature.