New install isn't idling correctly
-
I'm using a C2D on an intel brand board and 2 intel Gbit NICs, using PFSense 2.2 amd64. 1G RAM (2x 512M Sticks) There is nothing exciting about this machine. There are no packages installed.
I grabbed a spare HDD and installed 2.2 to that, and have the 2.1.5 HDD sitting to revert.
I did several backups of 2.1.5 (full, no RRD, no Packages, and no RRD & Packages).
I restored the "no packages" backup.
Everything else works the way I expect.The only slightly odd thing I did is move from i386 to AMD64. But it is a new install, not an upgrade.
I'm now seeing, load average above 1.0, elevated temps (52c, when before it was always sub 30c), the fan is spinning faster and it feels warmer. So, the machine is definitely working more.Using 2.1 through 2.1.5 I never had an issue with idle.
I've checked the BIOS, and the only setting that seems relevent is the IEST setting, which is enabled.
On the front page of the web interface the CPU Type is
Intel(R) Core(TM)2 CPU 4400 @ 2.00GHz 2 CPUs: 1 package(s) x 2 core(s)The speed with 2.1.5 varied down to 400Mhtz. It isn't with 2.2
The "top" command gives (And remember it's a Core 2 duo, so load average of 1 means 1 core is fully loaded)
$ top last pid: 98592; load averages: 1.01, 1.04, 0.87 up 0+00:25:00 20:52:59 28 processes: 1 running, 27 sleeping Mem: 41M Active, 29M Inact, 91M Wired, 1768K Cache, 69M Buf, 790M Free Swap: PID USERNAME THR PRI NICE SIZE RES STATE C TIME WCPU COMMAND 90963 root 1 20 0 219M 37824K piperd 1 0:00 0.39% php-fpm 15544 root 1 20 0 12464K 2172K select 1 0:01 0.00% apinger 12894 root 1 20 0 16812K 2344K bpf 0 0:00 0.00% filterlog 278 root 1 28 0 219M 23676K kqread 0 0:00 0.00% php-fpm 36717 root 1 52 20 17144K 2488K wait 1 0:00 0.00% sh 20363 root 1 20 0 46700K 5796K kqread 1 0:00 0.00% lighttpd 64604 root 1 20 0 28172K 18060K select 0 0:00 0.00% ntpd 44259 root 1 20 0 14548K 2060K select 1 0:00 0.00% powerd 46825 root 1 20 0 21160K 4360K select 1 0:00 0.00% miniupnpd 26974 nobody 1 23 0 30264K 4336K select 1 0:00 0.00% dnsmasq 6391 root 1 20 0 43612K 6300K select 0 0:00 0.00% mpd5 15565 root 1 20 0 28332K 3012K piperd 1 0:00 0.00% rrdtool 61695 root 1 52 0 43576K 2768K wait 1 0:00 0.00% login 98592 root 1 21 0 21996K 2812K CPU0 0 0:00 0.00% top 61844 root 1 52 0 17144K 2752K wait 1 0:00 0.00% sh 294 root 1 40 20 19032K 2528K kqread 1 0:00 0.00% check_reload_status 54472 root 1 20 0 16672K 2408K nanslp 1 0:00 0.00% cron 62108 root 1 52 0 17144K 2636K ttyin 1 0:00 0.00% shSystem activity gives
last pid: 11971; load averages: 1.25, 1.09, 0.90 up 0+00:26:52 20:54:51 80 processes: 8 running, 59 sleeping, 13 waiting Mem: 41M Active, 29M Inact, 91M Wired, 1768K Cache, 70M Buf, 790M Free Swap: PID USERNAME PRI NICE SIZE RES STATE C TIME WCPU COMMAND 11 root 155 ki31 0K 32K RUN 0 16:24 60.99% [idle{idle: cpu0}] 11 root 155 ki31 0K 32K RUN 1 16:07 40.97% [idle{idle: cpu1}] 0 root 8 0 0K 192K CPU0 0 6:29 35.99% [kernel{acpi_task_0}] 0 root 8 0 0K 192K RUN 0 6:58 34.86% [kernel{acpi_task_2}] 0 root 8 0 0K 192K RUN 1 6:46 34.57% [kernel{acpi_task_1}] 1486 root 24 0 219M 32780K RUN 0 0:00 0.49% php-fpm: pool lighty (php-fpm) 0 root -16 0 0K 192K swapin 1 0:36 0.00% [kernel{swapper}] 4 root -16 - 0K 32K - 0 0:02 0.00% [cam{scanner}] 0 root -92 0 0K 192K - 0 0:01 0.00% [kernel{em1 taskq}] 0 root -92 0 0K 192K - 1 0:01 0.00% [kernel{em0 taskq}] 12 root -60 - 0K 208K WAIT 0 0:01 0.00% [intr{swi4: clock}] 15544 root 20 0 12464K 2172K select 0 0:01 0.00% /usr/local/sbin/apinger -c /var/etc/apinge 5 root -16 - 0K 16K pftm 0 0:01 0.00% [pf purge] 12894 root 20 0 16812K 2344K bpf 0 0:00 0.00% /usr/local/sbin/filterlog -i pflog0 -p /va 278 root 28 0 219M 23676K kqread 1 0:00 0.00% php-fpm: master process (/usr/local/lib/ph 36717 root 52 20 17144K 2488K wait 1 0:00 0.00% /bin/sh /var/db/rrd/updaterrd.sh 20363 root 20 0 46700K 5844K kqread 1 0:00 0.00% /usr/local/sbin/lighttpd -f /var/etc/light 15 root -16 - 0K 16K - 1 0:00 0.00% [rand_harvestq]It's clear that something is pegging one core, but I"m stuffed if I can see what.
A possible, but probably not related problem, is I no longer see traffic grahps. But RRD Graphs work. I can see 50% CPU load there as well.
-
I think this was a case of co-incidence. Or the hardware didn't like having the HDD changed. But something jammed one core on, at 100% load and speed. Something low level, because the effect was also present just in the BIOS.
I swapped back to 2.1.5 and the issue remained.
I jiggered in the BIOS, but couldn't make any difference.
Eventually I did a full power off, pulled the power lead, tried to power back up (to drain the capacitors in the PSU) and then powered back up, and all is good in 2.1.5
I'll swap back to 2.2 in the near future.
So, probably not related to 2.2.
-
I strongly suspect I'm just talking to myself. But just for completion ….
I swapped back to the 2.2 AMD64 disk and the thing is correctly idling and changing speeds up and down as it should. So, just some stupidity in the BIOS.
It seems to run at a slightly higher base rate (ie, 450mhtz is the more common minimum now, than 300mhtz). But I put that down to a slightly higher base load in FreeBSD 10 over 8. Or unbound, or the new PHP or something else.
-
You definitely were chewing a good deal of CPU in the one post above, nearly all in something to do with ACPI from the looks of it.
0 root 8 0 0K 192K CPU0 0 6:29 35.99% [kernel{acpi_task_0}] 0 root 8 0 0K 192K RUN 0 6:58 34.86% [kernel{acpi_task_2}] 0 root 8 0 0K 192K RUN 1 6:46 34.57% [kernel{acpi_task_1}]It seems to run at a slightly higher base rate (ie, 450mhtz is the more common minimum now, than 300mhtz). But I put that down to a slightly higher base load in FreeBSD 10 over 8. Or unbound, or the new PHP or something else.
All else being equal, all of those things actually use less hardware resources in 2.2 than in 2.1.x versions. FreeBSD 10's packet filter is more efficient, other parts of the network stack are more efficient, in some cases drivers are better. Switching to PHP-FPM reduced its resource utilization. Unbound uses a trivial amount of resources.
If your CPU usage still seems high, monitor "top -SH" via SSH. Maybe something like the above acpi stuff going on. If there's a BIOS update available, installing it would probably be a good idea, as it might be some hardware bug that just wasn't exposed by FreeBSD 8 for some reason. Any time ACPI and "stupidity in the BIOS" come up in a thread, a BIOS update comes to mind.
-
You may also want to check your powerd settings in System > Advanced, Miscellaneous tab.
2.2 introduced a new power mode, 'unknown', and that defaults to HiAdaptive. If you were previously using Adaptive and the new power mode is what's detected then the box will average a higher CPU frequency.That clearly doesn't explain your >1 load average though.
Steve
-
The BIOS for such an old board (remember it's a core 2 duo system) is as up to date as it's going to get.
I had seen the unknown power management option, and now all are set the same, to adaptive.
It seems to run at about the same temperature, or maybe even slightly cooler. So maybe the displayed current CPU speed is calculated differently, between FreeBSD 8 & 10. I assume there is lots of scope for such a minor change to have been made.
