Re[8]: high load system do not take all CPU time
Здравствуйте, Robert.
Вы писали 26 декабря 2011 г., 23:54:59:
RB [drivelectomy -- 200+ lines]
RB You've been told the following, *repeatedly*:
RB Your hardware is not capable of keeping up with the level of network traffic
RB it is being subjected to.
RB Reaaltek cards and the 're' device driver are a *BAD*CHOICE* for systems
with
RB heavy network traffic. They're merely 'medium lousy' on a lightly-loaded
RB system, but you don't notice the problems under light loads.
RB You have two choices:
RB 1) live with the crappy performance
RB 2) get a better quality network card.
better card do not change situation http://youtu.be/f90nMtNdKB8
full video you can download
http://www.filehosting.org/file/details/316076/1.rar
igb3@pci0:1:0:3:class=0x02 card=0x00018086 chip=0x15218086 rev=0x01
hdr=0x00
vendor = 'Intel Corporation'
class = network
subclass = ethernet
12 root -32- 0K 624K WAIT1 1:38 0.00% intr{swi4: cloc
0 root -680 0K 384K - 3 1:08 0.00% kernel{dummynet
0 root -160 0K 384K sched 2 0:44 0.00% kernel{swapper}
12 root -44- 0K 624K WAIT3 0:31 0.00% intr{swi1: neti
12 root -44- 0K 624K WAIT3 0:11 0.00% intr{swi1: neti
12 root -68- 0K 624K WAIT2 0:07 0.00% intr{irq263: ig
12 root -44- 0K 624K WAIT2 0:06 0.00% intr{swi1: neti
13 root -16- 0K64K sleep 1 0:06 0.00% ng_queue{ng_que
13 root -16- 0K64K sleep 1 0:06 0.00% ng_queue{ng_que
13 root -16- 0K64K sleep 1 0:06 0.00% ng_queue{ng_que
13 root -16- 0K64K sleep 1 0:06 0.00% ng_queue{ng_que
12 root -68- 0K 624K WAIT3 0:05 0.00% intr{irq276: re
12 root -44- 0K 624K WAIT3 0:04 0.00% intr{swi1: neti
14 root -16- 0K16K - 0 0:04 0.00% yarrow
12 root -68- 0K 624K WAIT1 0:01 0.00% intr{irq262: ig
12 root -68- 0K 624K WAIT1 0:01 0.00% intr{irq257: ig
12 root -68- 0K 624K WAIT0 0:01 0.00% intr{irq261: ig
12 root -68- 0K 624K WAIT2 0:01 0.00% intr{irq258: ig
12 root -68- 0K 624K WAIT3 0:01 0.00% intr{irq274: ig
12 root -68- 0K 624K WAIT3 0:01 0.00% intr{irq264: ig
--
С уважением,
Коньков mailto:kes-...@yandex.ru
___
freebsd-questions@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-questions
To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
Re[8]: high load system do not take all CPU time
I get it!!
When one of netisr take 100% of CPU other netisr threads did not get
free CPU time.
http://piccy.info/view3/2444937/25e978a34d1da6b62e4e4602dee53d8b/
In this case network works without any problem
last pid: 23632; load averages: 5.53, 5.76, 5.72up
6+00:09:50 20:37:43
292 processes: 12 running, 265 sleeping, 15 waiting
CPU 0: 9.4% user, 0.0% nice, 18.4% system, 36.1% interrupt, 36.1% idle
CPU 1: 2.4% user, 0.0% nice, 12.2% system, 62.4% interrupt, 23.1% idle
CPU 2: 1.6% user, 0.0% nice, 4.3% system, 85.1% interrupt, 9.0% idle
CPU 3: 1.6% user, 0.0% nice, 3.9% system, 86.3% interrupt, 8.2% idle
Mem: 613M Active, 2788M Inact, 315M Wired, 122M Cache, 112M Buf, 59M Free
Swap: 4096M Total, 30M Used, 4065M Free
PID USERNAME PRI NICE SIZERES STATE C TIME WCPU COMMAND
12 root -72- 0K 160K CPU33 43.4H 100.00% {swi1: netisr
3}
12 root -72- 0K 160K CPU22 28.6H 93.60% {swi1: netisr 1}
12 root -72- 0K 160K CPU11 954:56 50.68% {swi1: netisr 2}
11 root 155 ki31 0K32K RUN 0 96.7H 36.91% {idle: cpu0}
12 root -72- 0K 160K RUN 0 757:29 31.10% {swi1: netisr 0}
11 root 155 ki31 0K32K RUN 1 98.4H 21.44% {idle: cpu1}
12 root -92- 0K 160K WAIT0 17.8H 12.94% {irq256: re0}
11 root 155 ki31 0K32K RUN 2 97.7H 11.08% {idle: cpu2}
11 root 155 ki31 0K32K RUN 3 98.6H 10.16% {idle: cpu3}
13 root -16- 0K32K sleep 0 411:30 4.25% {ng_queue0}
13 root -16- 0K32K sleep 0 411:54 4.20% {ng_queue3}
13 root -16- 0K32K RUN 0 411:05 4.10% {ng_queue1}
13 root -16- 0K32K RUN 2 411:16 3.81% {ng_queue2}
5588 root210 222M 106M select 0 116:59 0.93% {mpd5}
5588 root210 222M 106M select 1 0:00 0.93% {mpd5}
5588 root210 222M 106M select 3 0:00 0.83% {mpd5}
5588 root210 222M 106M select 0 0:00 0.83% {mpd5}
5588 root210 222M 106M select 3 0:00 0.83% {mpd5}
5588 root210 222M 106M select 3 0:00 0.83% {mpd5}
5588 root210 222M 106M select 3 0:00 0.83% {mpd5}
32882 root210 15392K 5492K select 1 313:35 0.63% snmpd
5588 root210 222M 106M select 3 0:00 0.54% {mpd5}
5588 root200 222M 106M select 0 0:00 0.15% {mpd5}
5588 root200 222M 106M select 2 0:00 0.15% {mpd5}
5588 root200 222M 106M select 3 0:00 0.15% {mpd5}
# netstat -W 1 re0
input(Total) output
packets errs idrops bytespackets errs bytes colls
96245 0 0 65271414 115828 0 80031246 0
104903 0 0 70367758 121943 0 85634456 0
102693 0 0 69663018 118800 0 83847075 0
108654 0 0 73776089 125368 0 88487518 0
100216 0 0 68186983 118522 0 80985757 0
94819 0 0 63001720 107334 0 73020011 0
108428 0 0 73849974 127976 0 88674709 0
# vmstat -i
interrupt total rate
irq14: ata0 13841687 26
irq16: ehci0 782489 1
irq23: ehci1 1046847 2
cpu0:timer2140530447 4122
irq256: re0374422787721
cpu1:timer2132118859 4106
cpu3:timer2108526888 4061
cpu2:timer2131292574 4105
Total 8902562578 17147
1 usersLoad 5.87 5.62 5.65 Jan 5 20:41
Mem:KBREALVIRTUAL VN PAGER SWAP PAGER
Tot Share TotShareFree in out in out
Act 737696 12284 315494038552 186192 count
All 946508 19628 540325288672 pages
Proc:Interrupts
r p d s w Csw Trp Sys Int Sof Flt760 cow 38031 total
227 128k 7771 14k 21k 13k 3996 1624 zfod 18 ata0 14
189 ozfod 1 ehci0 16
7.8%Sys 68.3%Intr 4.2%User 0.0%Nice 19.7%Idle 11%ozfod 2 ehci1 23
||||||||||| daefr 4123 cpu0:timer
++ 1439 prcfr 21603 re0 256
184 dtbuf 3275 totfr 4102 cpu1:timer
Namei Name-cache Dir-cache142271 desvn react 4083 cpu3:timer
Callshits %hits % 46214 numvn pdwak 4099 cpu2:timer
26140 24882 95
Re[8]: high load system do not take all CPU time
RB [drivelectomy -- 200+ lines] RB You've been told the following, *repeatedly*: RB Your hardware is not capable of keeping up with the level of network traffic RB it is being subjected to. RB Reaaltek cards and the 're' device driver are a *BAD*CHOICE* for systems with RB heavy network traffic. They're merely 'medium lousy' on a lightly-loaded RB system, but you don't notice the problems under light loads. RB You have two choices: RB 1) live with the crappy performance RB 2) get a better quality network card. without only one ipfw fw rule: queue 54 config pipe 54 queue 50 mask dst-ip 0x gred 0.002/10/30/0.1 275 queue 54 all from any not 80,110 to any in recv re0 works more! better: http://piccy.info/view3/2418620/59aa576c1006bbb046a13554d8468a6c/ with igb cards I get problems too! it put pptp traffice only to one queue0 instead to spread to all: queue0 queue1 queue2 queue3 =`( -- С уважением, Коньков mailto:kes-...@yandex.ru ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to freebsd-questions-unsubscr...@freebsd.org
