Pi hangs when nmea2000 connected
I have a pi5 with SSD and waveshare 2 channel isolated canhat, running a recent version of openplotter. Everything works great on my bench, connected to a garmin chartplotter and/or simulated data. I then sent it to a friend in New Zealand who installed it in his boat. Everything is good until he connects to his n2k network, then the cpu utilization shoots to 100% and the pi becomes unresponsive. Mouse stops working, doesn't respond to pings, no way to interact with it, but the display on hdmi is still okay. Power is provided by a U6223 buck converter that provides 6A, so I don't think it's a power problem. Maybe to much nmea data? Would filtering help? Any ideas on how to proceed or debug? Thanks!
9 Replies
I would be very surprised if the issue was too much data
More likely that there is some data coming in that is causing an issue
I would start by taking everything off the network and adding them back one at a time.
Or at least a segment of the network one at a time
Thanks, I'll try that!
I'm getting on a call with him shortly so we can try that. If I find the offending device, is there some way to filter out that data?
depend where the issue is
Appreciate your advice. Do you have any suggestions on next steps if I do isolate a device?
do they have any outgoing data connections? if they have is there a possibility of a data loop - the server getting its output back as input, and round it goes...a rather typical mechanism to get 100% cpu
disable all plugins to eliminate a plugin causing problems (when it gets real world data as opposed to what you had on your bench)
if you do isolate a device activate data logging for all input connections and create some logs: "healthy" without the offender, "trouble" with the offender active
..or even if you can't isolate an offender one step forward is to share a data log to see what data the server is busy processing
Yeah, turning off plugins was probably a better first step…
A good example would be if you are outputting n2k to 0183 and it ends up going to a 0183 to n2k converter. That would cause a loop.
Great advice, thanks! I'll dig into that and let you know
Okay, we tried disconnecting the most likely offenders (fusion radio and aftermarket engine monitor) with no improvement. Then turned off all plugins and tried again, still went to 100% and locked up. Next was going to capture debug data, but when we went to the Data Connections panel to enable debug, there is nothing listed there! This despite the fact that the Dashboard shows that can0 is connected. In fact, when it locks up, the dashboard shows 87.8 deltas/sec from can0. We tried re-adding can0 on the Data Connections panel, but still nothing shows up. Any idea what is going on with that? Screenshots attached
Never happened before. Maybe just rename the original .signalk directory and start from scratch