high cpu usage by system process
Hi peeps,
Recently i've noticed a significant slowdown on an old OpenBSD 3.5 box, which we run IPSec/VPN on. System load consistently reached 2.70+, and upon closer investigation it seems the system process hogs up the cpu.
We're using some el cheapo NICs on it, so i began to suspect that maybe the box is getting drowned in interrupts...but the systat reports below doesn't suggest that:
************************************************** ************************************************** **
load averages: 2.58, 2.50, 2.27 18:21:36
14 processes: 1 running, 13 idle
CPU states: 0.0% user, 0.0% nice, 96.2% system, 0.4% interrupt, 3.3% idle
Memory: Real: 4764K/60M act/tot Free: 439M Swap: 0K/4093M used/tot
PID USERNAME PRI NICE SIZE RES STATE WAIT TIME CPU COMMAND
4930 root 2 0 308K 992K idle select 0:00 0.00% sshd
5549 root 28 0 216K 860K run - 0:00 0.00% top
31061 root 2 0 292K 564K idle select 0:00 0.00% cron
5500 _syslogd 2 0 172K 496K idle poll 0:00 0.00% syslogd
23416 root 2 0 364K 1476K sleep select 0:00 0.00% sshd
1 root 10 0 368K 228K idle wait 0:00 0.00% init
11404 root 18 0 340K 304K idle pause 0:00 0.00% csh
22485 root 2 0 144K 456K idle netio 0:00 0.00% syslogd
31382 root 3 0 92K 476K idle ttyin 0:00 0.00% getty
13179 root 2 0 104K 492K idle select 0:00 0.00% inetd
19389 root 3 0 84K 476K idle ttyin 0:00 0.00% getty
10293 root 3 0 88K 476K idle ttyin 0:00 0.00% getty
19331 root 3 0 56K 476K idle ttyin 0:00 0.00% getty
20411 root 3 0 56K 476K idle ttyin 0:00 0.00% getty
************************************************** ************************
output of systat :io
/0 /1 /2 /3 /4 /5 /6 /7 /8 /9 /10
Load Average ||||||||||||||||||
/0 /10 /20 /30 /40 /50 /60 /70 /80 /90 /100
cpu user|
nice|
system|XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXX
interrupt|
idle|
/0 /10 /20 /30 /40 /50 /60 /70 /80 /90 /100
wd0 Kps|
tps|
************************************************** ************************
output of systat :swap
/0 /1 /2 /3 /4 /5 /6 /7 /8 /9 /10
Load Average ||||||||||||||
Disk 1K-blocks Used /0% /10% /20% /30% /40% /50% /60% /70% /80% /90% /100%
swap_ 4191248 0
************************************************** ************************
output of mbufs:
/0 /1 /2 /3 /4 /5 /6 /7 /8 /9 /10
Load Average >>>>>>>>>>>
/0 /5 /10 /15 /20 /25 /30 /35 /40 /45 /50 /55 /60
sockets XXX
data X
headers X
************************************************** ************************
output of pigs:
/0 /1 /2 /3 /4 /5 /6 /7 /8 /9 /10
Load Average <<<<<<<<<<<<<<
/0 /10 /20 /30 /40 /50 /60 /70 /80 /90 /100
<idle> XXXXXXXXXXXXX
************************************************** ************************
1 users Load 2.19 2.72 2.51 Thu Oct 7 18:27:26 2010
memory totals (in KB) PAGING SWAPPING Interrupts
real virtual free in out in out 8506 total
Active 4948 4948 449388 ops 100 clock
All 61624 61624 4640632 pages 4593 rl0
128 rtc
Proc:r d s w Csw Trp Sys Int Sof Flt forks 3685 vr0
2 56 7 44 13210 1 10 fkppw pciide1
fksvm
91.3% Sys 0.0% User 0.0% Nice 8.4% Idle pwait
| | | | | | | | | | | relck
============================================== rlkok
noram
Namei Sys-cache Proc-cache No-cache ndcpy
Calls hits % hits % miss % fltcp
13 13 100 1 8 -1 -8 1 zfod
cow
Disks wd0 128 fmin
seeks 170 ftarg
xfers itarg
Kbyte 54 wired
sec pdfre
Showing vmstat, refresh every 5 seconds.
************************************************** ************************
i'm beginning to suspect that maybe the box can't handle all that VPN/IPSec processing; what else can i check to determine this?
And how to do i solve this btw? Currently the box has 512MB RAM and a P4 2.0Ghz processor; i doubt if bumping up the RAM and going dual core is going to solve things...
tia
|