On Sunday, November 18, 2018 at 6:00:51 PM UTC-8, Steve2Q wrote: > > Hello again: after October 19, I had my new pi with a new install of weewx > and steel gauges running. I had another crash today after almost 11 days of > running and this is a snip from syslog which continues until a hard reboot > of the pi: Any thoughts? Could Steel Gauges be causing this. Also attached > are some results after the df command was issued. >
Well looking at the 'week' and 'month' cmon plots at http://photokinetics.org/Weather/cmon/ <http://www.google.com/url?q=http%3A%2F%2Fphotokinetics.org%2FWeather%2Fcmon%2F&sa=D&sntz=1&usg=AFQjCNGvqZgXrMkJLXOe4gp6MP3k4XcPyQ> you certainly seem to have quite a memory leak but there is no data in the plots saying what is leaking. Best I can suggest is to run "top" occasionally, perhaps once per day, and save the output and see if you can see who is using up memory. An example from my pi-zero: pi@zero:~ $ top top - 18:10:42 up 18 days, 6:27, 1 user, load average: 1.46, 1.34, 1.28 Tasks: 69 total, 2 running, 43 sleeping, 0 stopped, 0 zombie %Cpu(s): 95.5 us, 4.5 sy, 0.0 ni, 0.0 id, 0.0 wa, 0.0 hi, 0.0 si, 0.0 st KiB Mem : 443892 total, 53052 free, 142484 used, 248356 buff/cache KiB Swap: 102396 total, 102140 free, 256 used. 218624 avail Mem PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 4156 root 20 0 14916 10940 6452 R 48.9 2.5 7599:57 python 290 root 20 0 140692 105996 6488 S 48.6 23.9 2091:12 weewxd The 'python' is a UDP listener program I wrote that basically just idles listening for broadcasts from my WeatherFlow station and then sends the output to MQTT on a different pi. The next line for weewxd of course is weewx using the WeatherFlow UDP driver, listening for the same messages and saving to sqlite and using the vanilla Standard skin. The resident set size on mine 'does' creep upward slowly, but it's waaaay from being any kind of issue and my experience is it'll level out eventually. I've never seen what you've experienced, so you must have unique or unusual software somehow. I'd suggest you run top periodically from cron, saving to a file. Once per day should be enough since your system goes whacko so quickly. A crontab along the lines of the following should get it done (untested): 10 * * * * top -b -n 1 > /root/top.`date +%Y%m%d.%H%M%S` That'll create files /root/top.20181118.121002 and the like that sort nicely (corresponds to Nov-18-2018 12:10:02) -- You received this message because you are subscribed to the Google Groups "weewx-user" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
