Weird. After 50 days, CPU utilization jumped to 25 - 26%



  • After searching and reading some similar posts, mostly CPU load slowly climbing, thought I would report this one.

    I have two pfSense boxes both using Atom dual core on Jetway mini ITX boards. They work great by the way. I have been using them on different networks as a captive portals. For 50 days straight, no reboots, etc, the CPU utilization was around 1 - 2 %

    After 50 days, it jumped to 25 - 26%! It remained steady at this. Just all of a sudden, both boxes within hours of each other, like the time difference it took me to get them installed and started up some 50 days ago.

    I rebooted the units and of course CPU utilization is "normal" again at about 1 - 2%. These boxes are not highly loaded by any means. Throughput has remained pretty steady over the 50+ days and even now at 1 - 2% utilization, the load and packet transfer etc, is pretty minimal and pretty much the same.

    Just thought I'd drop this info in the fray in case someone has some ideas on why it would be doing this.

    John



  • Not easy to debug if it takes over 7 weeks for the "problem" to show up.

    Next time it happens, connect to the pfSense shell and give the command top -S to see which processes (including system processes) are using the most CPU time.



  • wallabybob,

    Thanks and yep…took a long time. Never once thought about doing a top. Getting old. I will next time, if there is a next time, and report back. Just thought I'd still report this bizarre behavior esp on two (2) boxes and at about the same time.

    John



  • johnjces,

    You said you have two pfSense boxes that cpu spiked at almost identical times. I assume these two are in fact on the same network? How did you happen to 'stumble' onto seeing the cpu spiking? Did you have a network slowdown,or? Wonder, if they are on the same network if your isp possibly was doing some re-routing stuff behind the scene and may have temporarily choked the routing tables to your pfsense's? Strange I know, but have seen adding routes in the wrong subnet for example cause some weirdness even to cpu/kernel whackyness.
    If these are on two different networks/locations would really be bizzare!

    Barry



  • I noticed something similar on my Nokia ip330 running nanobsd 1.2.3-RELEASE after about 5 weeks of uptime where CPU utilization spiked for a couple of weeks before returning to "normal".  When I ran "top -SH", the "idle" process was utilizing most of the CPU resources despite the fact that the aggregated data at the top of the page indicated overall "idle" CPU utilization was 0% but "nice" CPU utilization had spiked.  FWIW, I never noticed any performance issues while this was going on.

    I've attached an rrd graph from my system if anyone is interested.




  • @brcisna

    The two boxes are on different networks, different public IPs but same ISP. I did notice the spike when logging in to view some stats and check the logs. The network didn't appear to slow down, but our pipe to these boxes is really pretty small. But I do not think that any throughput suffered. A reboot cured the problem for now so if our ISP was doing 'something' it cured our little anomaly.


Locked