Re: dummynet: waking up pipe

2010-01-22 Thread Michael Spratt

you should use vstat or something to see if its interrupts eating your cpu..

Luigi Rizzo wrote:

On Fri, Jan 22, 2010 at 07:42:46PM +0300, Evgenii Davidov wrote:

,

On Fri, Jan 22, 2010 at 02:46:28PM +0100, Luigi Rizzo ?:


On Fri, Jan 22, 2010 at 04:35:35PM +0300, Evgenii Davidov wrote:
...

my problem is that dummynet cpu usage jumps from 0 to 99%:

   33 root1 -68- 0K 8K -   1 512.6H 86.18% dummynet
   33 root1 -68- 0K 8K -   1 512.6H 85.89% dummynet
   33 root1 -68- 0K 8K -   1 512.6H 88.28% dummynet
   33 root1 -68- 0K 8K -   0 512.6H 78.17% dummynet
   33 root1 -68- 0K 8K -   0 512.6H  0.88% dummynet
   33 root1 -68- 0K 8K -   1 512.6H  0.10% dummynet
   33 root1 -68- 0K 8K -   1 512.7H  0.00% dummynet
   33 root1 -68- 0K 8K -   1 512.7H  0.10% dummynet

i am not too clear on why this happens -- there shuold be only
one instance of the dummynet thread, not 8 as in your case.

no-no, one instance, it's a grep of a top log

ok then it's rather normal -- the task wakes up every millisecond and
it might have hundreds of packets to send at once, especially when
the machine is heavily loaded as in your case.

but since i restricted dummunet to run only on cpu 0 it's much better
maybe it's some smp problem in freebsd7


maybe the memory affinity helps a bit here.

cheers
luigi

--
Evgenii V Davidov
___
freebsd-ipfw@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-ipfw
To unsubscribe, send any mail to "freebsd-ipfw-unsubscr...@freebsd.org"

___
freebsd-ipfw@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-ipfw
To unsubscribe, send any mail to "freebsd-ipfw-unsubscr...@freebsd.org"



--
--

Mike Spratt - Iraq (GMT+3)

Office: (214)242-1782

DSN: (318)847-2983

Cell: +964-770-398-4366

m...@magicislandtechnologies.com

---
___
freebsd-ipfw@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-ipfw
To unsubscribe, send any mail to "freebsd-ipfw-unsubscr...@freebsd.org"


Re: dummynet: waking up pipe

2010-01-22 Thread Luigi Rizzo
On Fri, Jan 22, 2010 at 07:42:46PM +0300, Evgenii Davidov wrote:
> ,
> 
> On Fri, Jan 22, 2010 at 02:46:28PM +0100, Luigi Rizzo ?:
> 
> > On Fri, Jan 22, 2010 at 04:35:35PM +0300, Evgenii Davidov wrote:
> > ...
> > > > > > my problem is that dummynet cpu usage jumps from 0 to 99%:
> > > > > > 
> > > > > >33 root1 -68- 0K 8K -   1 512.6H 86.18% 
> > > > > > dummynet
> > > > > >33 root1 -68- 0K 8K -   1 512.6H 85.89% 
> > > > > > dummynet
> > > > > >33 root1 -68- 0K 8K -   1 512.6H 88.28% 
> > > > > > dummynet
> > > > > >33 root1 -68- 0K 8K -   0 512.6H 78.17% 
> > > > > > dummynet
> > > > > >33 root1 -68- 0K 8K -   0 512.6H  0.88% 
> > > > > > dummynet
> > > > > >33 root1 -68- 0K 8K -   1 512.6H  0.10% 
> > > > > > dummynet
> > > > > >33 root1 -68- 0K 8K -   1 512.7H  0.00% 
> > > > > > dummynet
> > > > > >33 root1 -68- 0K 8K -   1 512.7H  0.10% 
> > > > > > dummynet
> > > > 
> > > > i am not too clear on why this happens -- there shuold be only
> > > > one instance of the dummynet thread, not 8 as in your case.
> > > 
> > > no-no, one instance, it's a grep of a top log
> > 
> > ok then it's rather normal -- the task wakes up every millisecond and
> > it might have hundreds of packets to send at once, especially when
> > the machine is heavily loaded as in your case.
> 
> but since i restricted dummunet to run only on cpu 0 it's much better
> maybe it's some smp problem in freebsd7

maybe the memory affinity helps a bit here.

cheers
luigi
> -- 
> Evgenii V Davidov
> ___
> freebsd-ipfw@freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-ipfw
> To unsubscribe, send any mail to "freebsd-ipfw-unsubscr...@freebsd.org"
___
freebsd-ipfw@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-ipfw
To unsubscribe, send any mail to "freebsd-ipfw-unsubscr...@freebsd.org"


Re: dummynet: waking up pipe

2010-01-22 Thread Evgenii Davidov
Здравствуйте,

On Fri, Jan 22, 2010 at 02:46:28PM +0100, Luigi Rizzo пишет:

> On Fri, Jan 22, 2010 at 04:35:35PM +0300, Evgenii Davidov wrote:
> ...
> > > > > my problem is that dummynet cpu usage jumps from 0 to 99%:
> > > > > 
> > > > >33 root1 -68- 0K 8K -   1 512.6H 86.18% 
> > > > > dummynet
> > > > >33 root1 -68- 0K 8K -   1 512.6H 85.89% 
> > > > > dummynet
> > > > >33 root1 -68- 0K 8K -   1 512.6H 88.28% 
> > > > > dummynet
> > > > >33 root1 -68- 0K 8K -   0 512.6H 78.17% 
> > > > > dummynet
> > > > >33 root1 -68- 0K 8K -   0 512.6H  0.88% 
> > > > > dummynet
> > > > >33 root1 -68- 0K 8K -   1 512.6H  0.10% 
> > > > > dummynet
> > > > >33 root1 -68- 0K 8K -   1 512.7H  0.00% 
> > > > > dummynet
> > > > >33 root1 -68- 0K 8K -   1 512.7H  0.10% 
> > > > > dummynet
> > > 
> > > i am not too clear on why this happens -- there shuold be only
> > > one instance of the dummynet thread, not 8 as in your case.
> > 
> > no-no, one instance, it's a grep of a top log
> 
> ok then it's rather normal -- the task wakes up every millisecond and
> it might have hundreds of packets to send at once, especially when
> the machine is heavily loaded as in your case.

but since i restricted dummunet to run only on cpu 0 it's much better
maybe it's some smp problem in freebsd7

-- 
Evgenii V Davidov
___
freebsd-ipfw@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-ipfw
To unsubscribe, send any mail to "freebsd-ipfw-unsubscr...@freebsd.org"


Re: dummynet: waking up pipe

2010-01-22 Thread Luigi Rizzo
On Fri, Jan 22, 2010 at 04:35:35PM +0300, Evgenii Davidov wrote:
...
> > > > my problem is that dummynet cpu usage jumps from 0 to 99%:
> > > > 
> > > >33 root1 -68- 0K 8K -   1 512.6H 86.18% 
> > > > dummynet
> > > >33 root1 -68- 0K 8K -   1 512.6H 85.89% 
> > > > dummynet
> > > >33 root1 -68- 0K 8K -   1 512.6H 88.28% 
> > > > dummynet
> > > >33 root1 -68- 0K 8K -   0 512.6H 78.17% 
> > > > dummynet
> > > >33 root1 -68- 0K 8K -   0 512.6H  0.88% 
> > > > dummynet
> > > >33 root1 -68- 0K 8K -   1 512.6H  0.10% 
> > > > dummynet
> > > >33 root1 -68- 0K 8K -   1 512.7H  0.00% 
> > > > dummynet
> > > >33 root1 -68- 0K 8K -   1 512.7H  0.10% 
> > > > dummynet
> > 
> > i am not too clear on why this happens -- there shuold be only
> > one instance of the dummynet thread, not 8 as in your case.
> 
> no-no, one instance, it's a grep of a top log

ok then it's rather normal -- the task wakes up every millisecond and
it might have hundreds of packets to send at once, especially when
the machine is heavily loaded as in your case.
___
freebsd-ipfw@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-ipfw
To unsubscribe, send any mail to "freebsd-ipfw-unsubscr...@freebsd.org"


Re: dummynet: waking up pipe

2010-01-22 Thread Evgenii Davidov
Здравствуйте,

On Fri, Jan 22, 2010 at 02:39:05PM +0100, Luigi Rizzo пишет:

> On Fri, Jan 22, 2010 at 04:10:26PM +0300, Evgenii Davidov wrote:
> > ,
> > 
> > On Tue, Jan 12, 2010 at 02:05:39PM +0300, Evgenii Davidov ?:
> > 
> > > hello
> > > 
> > > when i enable net.inet.ip.dummynet.debug i get a lot of:
> > > 
> > > Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 1380 at 1460
> > > Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 2931 at 40
> > > Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 5189 at 40
> > > Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 4562 at 112
> > > Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 3610 at 51
> > > Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 597 at 40
> > > Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 1380 at 1400
> > > Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 1445 at 99
> > > Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 3610 at 51
> > > Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 3610 at 40
> > > Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 597 at 40
> > > Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 218 at 1500
> > > 
> > > is it bad?
> 
> this is just a debug message so ignore it.
> 
> > > let's see 3610:
> > > 
> > > ipfw pipe 3610 config bw 10500Kbit/s
> > > ipfw queue 3610 config pipe 3610 mask dst-ip 0x
> > > ipfw table 80 add 10.168.167.227 3610
> > > ipfw table 80 add 89.222.187.184 3610
> > > 
> > > 00040   8708035   7887386208 queue tablearg ip from any to table(80) out 
> > > via bge1
> > > 00050  44793529  40386675667 pipe tablearg ip from any to table(2) out 
> > > via bge1
> > > 00070   8006000   3713846496 pipe 16 ip from table(16) to any in via bge1
> > > 00071  68767107  23673242157 pipe 5 ip from not table(16) to any in via 
> > > bge1
> > > 
> > > ipfw pipe 3610 show
> > > 
> > > 03610:  10.500 Mbit/s0 ms   50 sl. 0 queues (1 buckets) droptail
> > > q03610: weight 1 pipe 3610   50 sl. 1 queues (512 buckets) droptail
> > > mask: 0x00 0x/0x -> 0x/0x
> > > BKT Prot ___Source IP/port Dest. IP/port Tot_pkt/bytes 
> > > Pkt/Byte Drp
> > >   5 ip   0.0.0.0/0  10.10.187.184/0 639563 40272017  0
> > > 0   0
> > > 
> > > my problem is that dummynet cpu usage jumps from 0 to 99%:
> > > 
> > >33 root1 -68- 0K 8K -   1 512.6H 86.18% 
> > > dummynet
> > >33 root1 -68- 0K 8K -   1 512.6H 85.89% 
> > > dummynet
> > >33 root1 -68- 0K 8K -   1 512.6H 88.28% 
> > > dummynet
> > >33 root1 -68- 0K 8K -   0 512.6H 78.17% 
> > > dummynet
> > >33 root1 -68- 0K 8K -   0 512.6H  0.88% 
> > > dummynet
> > >33 root1 -68- 0K 8K -   1 512.6H  0.10% 
> > > dummynet
> > >33 root1 -68- 0K 8K -   1 512.7H  0.00% 
> > > dummynet
> > >33 root1 -68- 0K 8K -   1 512.7H  0.10% 
> > > dummynet
> 
> i am not too clear on why this happens -- there shuold be only
> one instance of the dummynet thread, not 8 as in your case.

no-no, one instance, it's a grep of a top log

> 
> cheers
> luigi
> ___
> freebsd-ipfw@freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-ipfw
> To unsubscribe, send any mail to "freebsd-ipfw-unsubscr...@freebsd.org"

-- 
Evgenii V Davidov
___
freebsd-ipfw@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-ipfw
To unsubscribe, send any mail to "freebsd-ipfw-unsubscr...@freebsd.org"


Re: dummynet: waking up pipe

2010-01-22 Thread Luigi Rizzo
On Fri, Jan 22, 2010 at 04:10:26PM +0300, Evgenii Davidov wrote:
> ,
> 
> On Tue, Jan 12, 2010 at 02:05:39PM +0300, Evgenii Davidov ?:
> 
> > hello
> > 
> > when i enable net.inet.ip.dummynet.debug i get a lot of:
> > 
> > Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 1380 at 1460
> > Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 2931 at 40
> > Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 5189 at 40
> > Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 4562 at 112
> > Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 3610 at 51
> > Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 597 at 40
> > Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 1380 at 1400
> > Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 1445 at 99
> > Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 3610 at 51
> > Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 3610 at 40
> > Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 597 at 40
> > Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 218 at 1500
> > 
> > is it bad?

this is just a debug message so ignore it.

> > let's see 3610:
> > 
> > ipfw pipe 3610 config bw 10500Kbit/s
> > ipfw queue 3610 config pipe 3610 mask dst-ip 0x
> > ipfw table 80 add 10.168.167.227 3610
> > ipfw table 80 add 89.222.187.184 3610
> > 
> > 00040   8708035   7887386208 queue tablearg ip from any to table(80) out 
> > via bge1
> > 00050  44793529  40386675667 pipe tablearg ip from any to table(2) out via 
> > bge1
> > 00070   8006000   3713846496 pipe 16 ip from table(16) to any in via bge1
> > 00071  68767107  23673242157 pipe 5 ip from not table(16) to any in via bge1
> > 
> > ipfw pipe 3610 show
> > 
> > 03610:  10.500 Mbit/s0 ms   50 sl. 0 queues (1 buckets) droptail
> > q03610: weight 1 pipe 3610   50 sl. 1 queues (512 buckets) droptail
> > mask: 0x00 0x/0x -> 0x/0x
> > BKT Prot ___Source IP/port Dest. IP/port Tot_pkt/bytes Pkt/Byte 
> > Drp
> >   5 ip   0.0.0.0/0  10.10.187.184/0 639563 40272017  00 
> >   0
> > 
> > my problem is that dummynet cpu usage jumps from 0 to 99%:
> > 
> >33 root1 -68- 0K 8K -   1 512.6H 86.18% dummynet
> >33 root1 -68- 0K 8K -   1 512.6H 85.89% dummynet
> >33 root1 -68- 0K 8K -   1 512.6H 88.28% dummynet
> >33 root1 -68- 0K 8K -   0 512.6H 78.17% dummynet
> >33 root1 -68- 0K 8K -   0 512.6H  0.88% dummynet
> >33 root1 -68- 0K 8K -   1 512.6H  0.10% dummynet
> >33 root1 -68- 0K 8K -   1 512.7H  0.00% dummynet
> >33 root1 -68- 0K 8K -   1 512.7H  0.10% dummynet

i am not too clear on why this happens -- there shuold be only
one instance of the dummynet thread, not 8 as in your case.

cheers
luigi
___
freebsd-ipfw@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-ipfw
To unsubscribe, send any mail to "freebsd-ipfw-unsubscr...@freebsd.org"


Re: dummynet: waking up pipe

2010-01-22 Thread Evgenii Davidov
Здравствуйте,

On Tue, Jan 12, 2010 at 02:05:39PM +0300, Evgenii Davidov пишет:

> hello
> 
> when i enable net.inet.ip.dummynet.debug i get a lot of:
> 
> Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 1380 at 1460
> Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 2931 at 40
> Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 5189 at 40
> Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 4562 at 112
> Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 3610 at 51
> Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 597 at 40
> Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 1380 at 1400
> Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 1445 at 99
> Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 3610 at 51
> Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 3610 at 40
> Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 597 at 40
> Jan 12 13:53:32 r2 kernel: dummynet: waking up pipe 218 at 1500
> 
> is it bad?
> 
> let's see 3610:
> 
> ipfw pipe 3610 config bw 10500Kbit/s
> ipfw queue 3610 config pipe 3610 mask dst-ip 0x
> ipfw table 80 add 10.168.167.227 3610
> ipfw table 80 add 89.222.187.184 3610
> 
> 00040   8708035   7887386208 queue tablearg ip from any to table(80) out via 
> bge1
> 00050  44793529  40386675667 pipe tablearg ip from any to table(2) out via 
> bge1
> 00070   8006000   3713846496 pipe 16 ip from table(16) to any in via bge1
> 00071  68767107  23673242157 pipe 5 ip from not table(16) to any in via bge1
> 
> ipfw pipe 3610 show
> 
> 03610:  10.500 Mbit/s0 ms   50 sl. 0 queues (1 buckets) droptail
> q03610: weight 1 pipe 3610   50 sl. 1 queues (512 buckets) droptail
> mask: 0x00 0x/0x -> 0x/0x
> BKT Prot ___Source IP/port Dest. IP/port Tot_pkt/bytes Pkt/Byte 
> Drp
>   5 ip   0.0.0.0/0  10.10.187.184/0 639563 40272017  00   > 0
> 
> my problem is that dummynet cpu usage jumps from 0 to 99%:
> 
>33 root1 -68- 0K 8K -   1 512.6H 86.18% dummynet
>33 root1 -68- 0K 8K -   1 512.6H 85.89% dummynet
>33 root1 -68- 0K 8K -   1 512.6H 88.28% dummynet
>33 root1 -68- 0K 8K -   0 512.6H 78.17% dummynet
>33 root1 -68- 0K 8K -   0 512.6H  0.88% dummynet
>33 root1 -68- 0K 8K -   1 512.6H  0.10% dummynet
>33 root1 -68- 0K 8K -   1 512.7H  0.00% dummynet
>33 root1 -68- 0K 8K -   1 512.7H  0.10% dummynet

try: cpuset -l 0 -p 33

> 
> it does not depend on pps and traffic, i watch netstat -w1
> 
> os is 7.2-STABLE
> 
> thank you!
> 
> -- 
> Evgenii V Davidov
> ___
> freebsd-ipfw@freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-ipfw
> To unsubscribe, send any mail to "freebsd-ipfw-unsubscr...@freebsd.org"

-- 
Evgenii V Davidov
___
freebsd-ipfw@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-ipfw
To unsubscribe, send any mail to "freebsd-ipfw-unsubscr...@freebsd.org"