OpenMarine
KPLEX maxing out a core after a day - Printable Version

+- OpenMarine (https://forum.openmarine.net)
+-- Forum: OpenPlotter (https://forum.openmarine.net/forumdisplay.php?fid=1)
+--- Forum: Bug Reports (https://forum.openmarine.net/forumdisplay.php?fid=4)
+--- Thread: KPLEX maxing out a core after a day (/showthread.php?tid=1355)

KPLEX maxing out a core after a day - PaddyB - 2018-07-30

Anyone else experiencing this?  After about 24h kplex will sit at 100% , restarting kplex fixes it -

[Image: 1GlBvDg.png]

RE: KPLEX maxing out a core after a day - e-sailing - 2018-07-30

I think there was a thread where openplotter died after one day. The issue is between signal and kplex. If there is no nmea0183 sentence in 2 seconds, signal thinks the connection is down and starts a new one. Kplex creates new connections but never close any old connections. This happens until you got to many open tasks. The next problem is that the log file fills up until there is no space on the disk.
Solution give kplex a sentence minimum every 2 seconds. Or generate it.

RE: KPLEX maxing out a core after a day - PaddyB - 2018-07-30

(2018-07-30, 06:14 PM)e-sailing Wrote: I think there was a thread where openplotter died after one day. The issue is between signal and kplex. If there is no nmea0183 sentence in 2 seconds, signal thinks the connection is down and starts a new one. Kplex creates new connections but never close any old connections. This happens until you got to many open tasks. The next problem is that the log file fills up until there is no space on the disk.
Solution give kplex a sentence minimum every 2 seconds. Or generate it.

Wonder if this is related..  There's no serial going into kplex but there is nmea data going in from signalk and some from node red. It all keeps working, but kplex just sits at 100% on one of the cores and everything slows down a bit. There are a load of  kplex threads on htop. ..

[Image: OYbpcjp.png]
Actually digging a bit deeper I think you're probably right, Esailing. I have raspbian node red and signalk node red doing stuff and it's like Westworld, I think it's on the verge of AI consciousness Wink
So looks like nothing was going through kplex - signalk node-red has been making a MDA nmea sentence so that opencpn logbook gets pressure data but this doesn't look like it goes through kplex. But it must, as that's the only place the Opencpn can get the data - from TCP port 10110 . But the sentence doesn't show up in diagnostic in any of the kplex entries. So dunno how that works?
Anyway, dummy sentence going from node-red into kplex now to see if that cures it.

So how does nmea get from a signalk node-red signalk-send-nmea node to tcp port 10110?

RE: KPLEX maxing out a core after a day - e-sailing - 2018-07-30

The port 10110 comes from signalk not from kplex. Check if your sentences come when you do "sudo killall kplex"

RE: KPLEX maxing out a core after a day - PaddyB - 2018-07-30

(2018-07-30, 09:07 PM)e-sailing Wrote: The port 10110 comes from signalk not from kplex. Check if your sentences come when you do "sudo killall kplex"

Thanks, had to resort to the manual ;Winkll makes sense now, I didn't realize nmea0183 went through signalk after kplex. Fingers crossed kplex will be back to normal now it has something to do . 

[Image: assets%2F-L9g5MoJ9GSCaNGq_yjb%2F-LCPrzzc...3d60d8b8e5]

RE: KPLEX maxing out a core after a day - PaddyB - 2018-07-31

Overnight and kplex still happy - prob solved by the looks of it. Thanks!  Cool