So, then I fired up Aurora to get the cob webs out of the machine/cable/etc. I went to start the sho and only network 2 did anything.

I checked the breakers, good.

I checked HW Util, good.

I tried the sequence again, only net 2.

I removed/readded the networks, only data on net 2.

I chaned com 4 to com 6, and updated Aurora, still only net 2.

I tried swaping the USB ports, only net 2.

 

Somehow, after cables, networks, coms, changing everything I got data on BOTH net and net2, I let Aurora run the full sone, and set it to loop.

When the song restarted after ending with both networks working, I only had net 2 again???

 

What's going on?

 

Edit...

 

It would be usefull if I gave a bit on other info, wouldn't it?

I'm on Windows 2003

Both adapters are D-Light V3.

RS485 driver is the existing one from last year.

Aurora v1.0.69

Logged in ad admin.

Last year's Aruora was .41?, I did an uninstall, then reboot, then install.

Hardware util is 2.4.1

 

Edit again...

 

So at the end of the last test, I shut everything down. Then I though... hum... I wonder if it just needed one more reboot.

Well, The intro ran the first time correctlty, then when it looped again, I only had net 2!

 

What the frig?

I never changed any of that from last year, always want with the default.

I can try 19200 if you tink that do something usefull. I'll have to set that in the HW util, and Auora, right?

 

 

56000 or 115200 are the best speeds to run depending on your hardware.  D-Light is now standardizing on 115200, so if you are using new firmware on all of the controllers then 115200 should be fine (although 56000 is still supported if you run into issues at 115200).  Some older firmwares have some glitches at that rate, but those have been long gone for a while now.  LOR used to require 56000, but they have supported 115200 for a couple years now without issue.  19200 works for debugging, but I really do not recommend it as at that slow speed it is really easy to congest the pipe.

 

Are your controllers on network 1 receiving heartbeat (solid LED instead of blinking LED) when they are not responding to commands?

 

Solid light on first loop for net1 (when responding),  blinking there after for net1 (when not responding).

Originally Com4 = Net1 & Com5 = Net2 (Net1 drops connection)

Currently Com4 = Net2 & Com5 = Net1 (Net1 drops connection)

 

Also, I tried moving from Com4 to Com6, but that did not change anything.

Network speeds were at 115K.

It sounds like the port isn't releasing for some reason and Aurora thinks it is still in use.  I know it sounds crazy, but I'm trying to record the exact failure mode here:

 

What happens if you change Net 1 to Net 3 (this would give you a Net 2 and 3, but no Net 1)?

Well, that is NOT what I had expected to happen!

With Com4 = Net3, Com5 = Net2 communication is stable and reliable though multiple loops. Everything works correctly.

Now, I did use a V2 adapter last year, but I don't remember what Com port was used. I do see that in the Computer Manager that Com3 is concidered "In Use" though. So last year Net1 was a different adapter and possibly a different Com port. That was before the uninstall/install of 1.0.69

 

I did move two channels over to Net3 to make sure that there was some data going over the network.

currently there is one physical controllers on each network (ACx16), and thre are six ACx16s configured for each network.

I added a single channel to Net1 Controler1 channel 1, then I added Net1 IDE0 (something high), then Net4/4/1, then Net 4/E6/1.

 

Once the network is lost, Aurora does need to be closed out, then restarted to get it back.

 

I also just tried a install/repair, got a message about msxml being locked.

Now Net2 that keps dropping.

 

I suppose that I could go back to how it was, then create Net3, then slowly start moving chennels until it fails again.

Took me a bit loner to get back to you this time because I was doing the last few bits of wireing of the new controllers. So....

Now I'm on the tablet, Windows XP, clean install as of 2009, up to date on all MS updates, clean install of 1.0.69, never had an old install.

I'm using the same two D-Light RS485 V3 adapters, D-Light's USB to USB(mini), known good ethernet cables.

 

Running burn-in sequece instead of show intro.

Different computer.

Brand new controllers.

OEM Firmware 1.14

Two controllers on each network (16/19 on Net1, 17/18 on Net2)

Net1 = Com4

Net2 = Com5

Entire song did not play though this time, I selected a 30 second timeframe to loop.

Speed set to 115K

 

 

Hmmm... I think I can see in the code how this could happen in Aurora, but I do not believe it should affect Borealis during a real show.  You might want to reboot and check Borealis to make sure I'm correct though.

 

Although I can't replicate it here, I believe each loop iteration is rapidly closing/reopening the COM port (instead of holding it open like it should) and the COM driver on your system isn't keeping up.  More specifically, the driver is writing commands from the buffer to the line when the close request comes in (so the buffer has to finish dumping before the close completes), but the open request arrives before the buffer is emptied.  I've seen this situation generate a race condition in the FTDI COM Emulation drivers before.  It puts the adapter into a state where it isn't open, but it isn't closed either.  Sometimes replugging the adapter (with Aurora closed) will reset it, but other times a full reboot is required.

 

We have a fix for the race condition in beta (it was developed as a part of 1.2), but it is far to extensive to successfully roll out for this season.  I'm looking at a simpler patch for this specific issue (improve the COM reset logic so the line holds during looping).  I have a note which says the current reset was put in place due to a Vista issue, so I may have to do a bit of testing to figure out what is going on with Vista before fixing this.

 

Alright, I'll create a simple show tomarow and see if on song 2, or loop 2 anything gets dropped. If Borealis is happy, then we're happy.

 

Can't replicate it? Weird. I have the exact same thing on two compleately seperate computers. Tablet's XP and a few years old though, may be slower and not communicating as fast. I think both computers are USB 1.0 though.

 

Although I have not tried Borealis on the show server, on my tablet after each song, the solid light blicked twice, then went solid for the next song. The same thing for when the show looped. I started and stopped the show a few times, all went well on the tablet. I'd expect the same for the show server. I'll test it on the server ths weekend. :)

 

The net dropping is obnoxious for testing, but not show stopping.

 

Seems like it may be a good idea to do a socket open on run/start, then close on end, just leave it open until a stop event. Also, possibly close/clear on application start? Not that I'm trying to tell you how to work. -- Just a thought I had.