High CPU load when some external devices connected via serial are not switched on

[UserMacUseface]

Hello,
i recently noticed that my raspberry pi 4 running openplotter with signalK was on high CPU load.
When checking processes via "top" I noticed the "node" process sitting at around 300.

In that particular case three external devices ( ais, depth sounder and vhf ) were switched off.
As soon as I switched these on the CPU usage dropped down to normal levels.

When checking the signalK server log I noticed a lot of entries reading:
bind EADDRINUSE 169.254.133.250:2053 at node:dgram:359:20 at processTicksAndRejections (node:internal/process/task_queues:84:21) { errno: -98, code: 'EADDRINUSE', syscall: 'bind', address: '169.254.133.250', port: 2053 }

Any ideas what is going on and how to prevent this from happening?

Thanks a lot

[tkurki]

The error is probably unrelated, but worth solving first to rule it out. It seems something is occupying UDP port 2053, could be some other application or if you have multiple connections you could have a duplicate clash, so that one connection grabs the port and another gives the error.

Steps to debug

• review your connections in SK and make sure there are no duplicates
• lsof -PiUDP will list all open UDP sockets

[UserMacUseface]

Hello.
I checked connections and there are no duplicates.
What info do you need from the lsof output?
thanks

Edit:
It seems I found the culprit for the probably unrelated error.
Removed the service and I do not get the mentioned error anymore.
However the CPU load problem is still there and a main issue.

[UserMacUseface]

Any further ideas where I could look? Thanks a lot.

[tkurki]

• Are the 3 devices NMEA0183 over serial connections? If yes does enabling signalk:streams:serialport debug key produce any meaningful output?
• Does restarting the server (with the devices off) result in the same CPU load?
• Try disabling all plugins to rule out a plugin doing something funny when there is no data. If that solves the issue enable plugins one by one to sort out the culprit

[UserMacUseface]

Ah! Thanks. That did help.
There was an error message about a USB-GPS device not being present. After disabling this device in the Data Connections. The CPU-Load did drop down to 5%
So maybe the fact that switching on some of the other devices and the cpu load dropping was just some coincidence.
Good to know the debug key in case of further issues.
Is it wise to disable the entry in the debug file for better performance, or does it not matter?
Thanks again

[tkurki]

Yes, debug output does increase cpu usage. Naturally depends on the frequency: per incoming n2k message is a lot different from some lines once at server startup.

I am fairly surprised at the cpu for a nonexistent connection, it should retry only once every while. But glad you got rid of the pest.