> Date: Mon, 3 Aug 2015 19:43:58 -0300
> From: [email protected]
> To: [email protected]
> Subject: Re: [pfSense] 2.2.4 high cpu usage
>
> On Sun, 2 Aug 2015 21:30:12 -0300
> Nenhum_de_Nos <[email protected]> wrote:
>
> > Hi,
> >
> > I was trying the 2.2.4 i386 nanobsd version on the soekris 5501-70. It
> > works better then the earlier 2.2.x releases, but yet there is a catch.
> >
> > My load quite always close to 1, when I have all two links full using one
> > conection on each (iperf test):
> >
> > last pid: 16151; load averages: 1.06, 1.37, 1.40 up 0+00:46:02
> > 00:17:03
> > 43 processes: 3 running, 40 sleeping
> > CPU: 47.6% user, 0.0% nice, 17.8% system, 24.5% interrupt, 10.0% idle
> > Mem: 39M Active, 43M Inact, 77M Wired, 60M Buf, 312M Free
> > Swap:
> >
> > PID USERNAME THR PRI NICE SIZE RES STATE TIME WCPU COMMAND
> > 16220 root 1 32 0 17248K 8304K RUN 7:24 15.87% lighttpd
> > 285 root 1 22 0 57384K 34516K accept 1:12 3.17% php-fpm
> > 43530 root 1 20 0 11384K 2572K RUN 0:01 0.20% top
> > 284 root 1 52 0 57384K 34476K accept 1:28 0.00% php-fpm
> > 283 root 1 52 0 57256K 33424K accept 1:14 0.00% php-fpm
> > 42648 root 1 20 0 12216K 4084K select 0:10 0.00% miniupnpd
> > 66652 unbound 1 20 0 21464K 11120K kqread 0:08 0.00% unbound
> > 10161 root 1 20 0 10356K 1968K bpf 0:03 0.00% filterlog
> > 11835 _dhcp 1 20 0 10312K 2044K select 0:02 0.00% dhclient
> > 34749 root 1 20 0 10124K 1900K select 0:02 0.00% apinger
> > 48046 root 1 20 0 10292K 2044K select 0:01 0.00% syslogd
> > 70277 root 1 52 20 10584K 2224K wait 0:01 0.00% sh
> > 315 root 1 20 0 8976K 3268K select 0:01 0.00% devd
> > 11863 root 1 20 0 13104K 13132K select 0:01 0.00% ntpd
> > 70222 root 1 20 0 10948K 3116K pause 0:00 0.00% tcsh
> > 282 root 1 20 0 53028K 18568K kqread 0:00 0.00% php-fpm
> > 35466 root 1 20 0 11572K 2632K piperd 0:00 0.00% rrdtool
> > 49692 root 1 20 0 10236K 1960K nanslp 0:00 0.00% cron
> >
> > This above is no web access. If I have a browser accessing the webui it get
> > worse a bit. The UI get slowish.
> >
> > When I stop the iperf's, even though no links usage (close to 0), the load
> > drops to more then I would imagine and the web server process still is
> > oscilating the cpu usage. Is this normal ? Sometimes it gets close to 7%
> > then back and forth to around 15-20%.
> >
> > last pid: 22953; load averages: 0.42, 0.57, 0.91 up 0+00:56:22
> > 00:27:22
> > 40 processes: 2 running, 38 sleeping
> > CPU: 33.8% user, 0.0% nice, 1.1% system, 2.2% interrupt, 62.8% idle
> > Mem: 37M Active, 44M Inact, 76M Wired, 60M Buf, 313M Free
> > Swap:
> >
> > PID USERNAME THR PRI NICE SIZE RES STATE TIME WCPU COMMAND
> > 16220 root 1 30 0 17248K 8304K RUN 8:58 13.67% lighttpd
> > 283 root 1 23 0 57256K 33476K accept 1:28 1.66% php-fpm
> > 43530 root 1 20 0 11384K 2632K RUN 0:04 0.29% top
> > 284 root 1 52 0 57384K 34476K accept 1:28 0.00% php-fpm
> > 285 root 1 52 0 57384K 34516K accept 1:16 0.00% php-fpm
> > 42648 root 1 20 0 12216K 4088K select 0:11 0.00% miniupnpd
> > 66652 unbound 1 20 0 21464K 11240K kqread 0:09 0.00% unbound
> > 10161 root 1 20 0 10356K 1968K bpf 0:04 0.00% filterlog
> > 11835 _dhcp 1 20 0 10312K 2044K select 0:02 0.00% dhclient
> > 34749 root 1 20 0 10124K 1900K select 0:02 0.00% apinger
> > 48046 root 1 20 0 10292K 2044K select 0:02 0.00% syslogd
> > 70277 root 1 52 20 10584K 2224K wait 0:01 0.00% sh
> > 315 root 1 20 0 8976K 3268K select 0:01 0.00% devd
> > 11863 root 1 20 0 13104K 13132K select 0:01 0.00% ntpd
> > 282 root 1 20 0 53028K 18568K kqread 0:00 0.00% php-fpm
> > 70222 root 1 20 0 10948K 3116K pause 0:00 0.00% tcsh
> > 35466 root 1 20 0 11572K 2632K piperd 0:00 0.00% rrdtool
> > 49692 root 1 39 0 10236K 1960K nanslp 0:00 0.00% cron
> >
> >
> > Is it safe to say the 2.2.x are good for those old soekris ?
> >
> > thanks,
> >
> > matheus
>
> Same old problem. When I tried to access the WEBUI, crash.
>
> So, no 2.2.x for 5501 ...
>
> matheus
Man who is sucking that many userland CPU cycles? Is it really lighttpd, up to
40%? And why is it affected under high network thoughput? Makes no sense from
the top output, does "top -PSH" helps suggest where all CPU is going to?
_______________________________________________
pfSense mailing list
https://lists.pfsense.org/mailman/listinfo/list
Support the project with Gold! https://pfsense.org/gold
