OpenCPN 5.6.2 - SignalK connection causes freeze

[mhember]

Dave,

No problem - in that case, have you got a machine I can ssh into to create an ssh tunnel for you? If not I can try with a machine I have running at home....

Alternatively, let me know how to dump some data stream for you?

Marcus

[jongough]

I have attached 2 files, both signalk delta files that 'should' demonstrate the issue. They can be replayed through signalk using a data connection with a data type of FileStream and type of Signal K.



I have tried it on my Pi 3B+ and it does cause the lockup for me, but.... It is rather uncontrollable as I am not really sure how to throttle it to a more reasonable speed.


I have just updated to the latest node.js, npm and signal k to see if that helps but to no avail.


If you use the TCP port 10110 for the nmea0183 data it works, well does not lock up, but the signalk port of 3000 will cause the issue.

[jongough]

As a test I copied opencpn 5.2.4 executable from a working Pi 3B+ SDcard onto the one I am using. I renamed the 5.6.2 binary and ran the 5.2.4 binary to see what would happen. With everything the same the 5.2.4 ran fine and 5.6.2 hung. I have attached a log file with both runs in it. The first is 5.2.4, the second is 5.6.2. Hope this helps.

[bdbcat]

Jon...
"I have attached 2 files, both signalk delta files that 'should' demonstrate the issue. They can be replayed through signalk using a data connection with a data type of FileStream and type of Signal K. "


Could you give me a tutorial on how to do that? Save me some time and aggravation...
I have sK server running on OpenPlotter.



Thanks
Dave

[jongough]

To help I think this page should give you the information you need:
https://pysselilivet.blogspot.com/20...enplotter.html


I have just installed the signalk-data-logger and signalk-log-player plugins into signalk. This allows writing all signalk data to a log file and then replaying it when needed. Do be mindful that you need to restart signalk everytime you enable or disable the plugin. I have noticed, on my system, that sometimes I have to restart the Pi as node.js can get a little resource hungry when changing things.


I have attached the log file I captured that you can play back to see what is happening.


If you use 5.2.4 it will work and display AIS targets and GPS info. If you use it with 5.6.2 it will work for ~10-30secs and will then lock up.


Hope this helps

[bdbcat]

Jon...
Thanks for the quick-start.
Running now, no problems seen so far on Windows or linux. I see a bunch of AIS targets in eastern AUS.


Is your sK client Windows, or other? Is your sK server localhost, or on other network accessible machine?
Dave

[jongough]

I am currently near brisbane so the files I sent you are for there. I am currently stationary, so you should see the other boats but not a lot else as I didn't want to clag up the system with data that does not cause an issue. I run opencpn on a raspberry Pi 3B+ using openplotter. I don't have a windows machine, virtual or otherwise, with me. If I don't use AIS data but do use wind and auto pilot it seems to be OK, so I suspect that the processing of the AIS data is too much for the pi 3B+ armhf engine and it goes into a loop.



For me ocpn hangs very quickly with 5.6.2, but runs fine with 5.2.4 (just different executable). Using TCP 10110 input it is OK, but processing signalk on TCP 3000 input appears to be the issue. I will try running the same test on an old Asus machine with Linux and see what happens.

[jongough]

I have now connected my boat pc running linux mint xfce4 una (20.3) to the raspberry pi 3B+ using openplotter 2.9.1 (ubuntu buster/10) both running opencpn 5.6.2. They both have the same issue with using signalk data to TCP port 3000. Here is the log from the asus PC (boat).

12:28:34.230 MESSAGE SignalKDataStream.cpp:328     WebSocket SignalKDataStream watchdog timeout: Signal K:10.10.10.1:3000
12:28:34.430 MESSAGE SignalKDataStream.cpp:610 Opening Signal K WebSocket client: Signal K:10.10.10.1:3000
12:28:34.445 MESSAGE SignalKEventHandler.cpp:65 Connected to Signal K server version: 1.43.0
12:28:54.448 MESSAGE SignalKDataStream.cpp:328     WebSocket SignalKDataStream watchdog timeout: Signal K:10.10.10.1:3000
12:28:54.648 MESSAGE SignalKDataStream.cpp:610 Opening Signal K WebSocket client: Signal K:10.10.10.1:3000
12:28:54.669 MESSAGE SignalKEventHandler.cpp:65 Connected to Signal K server version: 1.43.0
12:29:02.656 MESSAGE SignalKDataStream.cpp:328     WebSocket SignalKDataStream watchdog timeout: Signal K:10.10.10.1:3000
12:29:02.656 MESSAGE SignalKDataStream.cpp:610 Opening Signal K WebSocket client: Signal K:10.10.10.1:3000
12:29:12.664 MESSAGE SignalKDataStream.cpp:328     WebSocket SignalKDataStream watchdog timeout: Signal K:10.10.10.1:3000
12:29:14.788 MESSAGE SignalKDataStream.cpp:610 Opening Signal K WebSocket client: Signal K:10.10.10.1:3000
12:29:15.160 MESSAGE SignalKEventHandler.cpp:65 Connected to Signal K server version: 1.43.0
12:30:00.018 MESSAGE chart1.cpp:7389 LOGBOOK:  2022-06-30 02:30:00 UTC  GPS Lat  -27.45444 Lon  153.18893 COG    0.00000 SOG   0.00 kts

Both go to 100% of one processor which effectively locks everything up until ocpn is closed (running from a terminal helps with this as it cannot be closed down by clicking on the 'X' button).

[jongough]

I tend to run a single PC (Pi 3B) when moving (sailing) due to power requirements. In dock, like now, I can run more as we have shore power. However I only use linux, mint for the main PC, openplotter ubuntu on the Pi.


So the issue happens on ubuntu 10 (buster). I would like to test ubuntu 11, but not sure it will be much different.

[bdbcat]

Jon...
One more question.
Are you running OCPN on both machines? If so, what happens if you run OCPN only on ASUS, and just running the sK server on the rPi? Any difference?


Thanks
Dave

[jongough]

I have defined nmea0183 udp connections for all data using signalk plugins for AIS and non-AIS data. Using this across a network connection it all works. It would appear that it is signalk process within ocpn that has the issue.

[jongough]

For the comms test the Pi ran signalk and the Asus ran OCPN 5.6.2. With the signalk stream coming in on 3000 (json messages) it locked up like the Pi.

[Hakan]

For info FWIW:
I've run 5.6.2 for weeks using the signalk protocol. Both on Win/Rpi clients and OCPN on the same RPi as the SK server. (About 30 deltas/sec and 50 paths. Including a "normal" cruising ship's data and more than 300 AIS targets)
I can see one difference though: I don't use the Openplotter 10.10.10.1 connection but instead a router subnet: 192.168.1.xx:3000. (I've some haze memory the 10.10.10.1 net is not stabel? The wifi part needs a special driver?)

[bdbcat]

Hakan...
That could be related.
I am having no trouble with linux, using a direct direct subnet address.


How do I activate the 10.10.10.1 ip address in the server side?


Dave

[jongough]

I know that using the Pi as a hotspot and connected as a router, i.e. two uses of the wifi at the same time, can be a problem. For most of the time, except when trying the stream on a different machine with the Pi getting and sending the data, I use local addresses, i.e. 127.0.0.1, as either a raw address or as 'localhost'. I moved to 127.0.0.1 as the pi kept on giving me ipv6 addresses and switching. It does not now.

The fact that OCPN 5.2.4 binary works and 5.6.2 doesn't, with that being the only change suggests that there is something different about the binaries and the way they process the messages. For me 5.6.2 works for a very short time then takes 100% of the processor and cannot be interacted with, i.e. it locks up and does not update the screen. The only reasonable way out of it is to terminate the process.

The test I did yesterday was to use the Pi to collect the data and then forward it to a Asus eeepc running OCPN. It locked up the same way when using signalk data, i.e. json messages, but worked fine with nmea0183. For the nmea0183 I had to use the udp broadcast of the nmea messages.

I have been running openplotter for a long time and this is the first time I have had this issue. I cannot debug the process I as I don't have access to anything to do that.

I can run signalk to get all the data and use the tcp 10110 port to pass this to OCPN. Using signalk json messages causes the lock up. Is there a version of software that OCPN uses that may cause the issue?

What information can I provide to help debug this issue?