cbq.init problem

[koks]

Pozdrav svima!
Imam problem sa podesavanjem cbq.init skripte (verzija 0.7.3, ako ima veze). Naime, internet delim na 4 racunara i sa ogranicavanjem downloada nema nikakvih problema, sve radi odlicno. Medjutim, problem je u ogranicavanju uploada. Mreza izgleda ovako:

internet -----> AP ----> router ------> lokalna mreza

Na internet se kacim preko wireless-a koji je uboden u linux router na eth0. Ostatak mreze (gde su 4 masine) je isto uboden u router na eth1 (192.168.1.0). Hocu da ogranicim upload na masini 192.168.1.30 na 30kbit/s. Mislim da sam sve podesio kako treba rtfm:, po uputstvu), ali mozda mi nesto promice. U nastavku je class fajl kojim ogranicavam upload:

DEVICE=eth0,100Mbit,10Mbit
BOUNDED=yes
PRIO=5
WEIGHT=3
RULE=192.168.1.30,
RATE=30kbit

Pomagajte! :crash:

 

[Sale99]

dodaj jos jedan red?

DEVICE=eth0,100Mbit,10Mbit
BOUNDED=yes
PRIO=5
WEIGHT=3
RULE=192.168.1.30,
RULE=,192.168.1.30
RATE=30kbit

 

[koks]

Na zalost, ni to nije pomoglo!
Ne znam da li ima veze, izlaz na net ide preko ppp0, ali fizicki na ruteru je to eth0 (iz eth0 ide kabel u AP). Koliko sam procitao cbq.init radi samo na ethernet adapterima.
Zbog ocaja, probao sam da stavim i ppp0 kao adapter koji shapeujem i (naravno) nije imalo uticaja. Koristim zipslack 10.2 (ako ima neke veze). :wall: :crash: :cuss:

 

[Paradigma]

DEVICE=eth1,100Mbit,10Mbit
RATE=30Kbit
WEIGHT=3Kbit
BOUNDED=yes
PRIO=5

RULE=192.168.1.30/32,0.0.0.0/0

 

[koks]

Ne pomaze ni to...
Da nije problem u NATovanju ili necemu slicnom?

Poz

 

[ravni]

tc -s qdisc ls dev eth0

Otkucaj to u konzoli pa paste-uj ovde

 

[koks]

Evo ga:

qdisc cbq 1: rate 2000Kbit (bounded,isolated) prio no-transmit
Sent 81468 bytes 178 pkts (dropped 0, overlimits 0)
borrowed 0 overactions 0 avgidle 3173 undertime 0
qdisc tbf 6: parent 1:6 rate 30000bit burst 10Kb lat 1.7s
Sent 0 bytes 0 pkts (dropped 0, overlimits 0)

Nadam se da pomaze...

 

[ravni]

to je ceo listing?

otkucaj i

tc -s class ls dev eth0

pa pasteuj ovde

Kako izgleda, nista ti ne prolazi kroz izlazni queue od 30K

moguce je da treba filter da namestis
(tipa tc filter add dev eth0 protocol ip parent 1:0 prio 1 u32 match ip src 192.168.1.30/32 flowid 6
mada ne znam zasto skripta to ne radi automatski, a i nekako mi je sumnjivo to 6: kod tbf qdisc-a

 

[koks]

Kada otkucam kao sto si rekao:

tc -s class ls dev eth0

dobijem sledece:

class cbq 1: root rate 2000Kbit (bounded,isolated) prio no-transmit
Sent 0 bytes 0 pkts (dropped 0, overlimits 0)
borrowed 0 overactions 0 avgidle 3275 undertime 0
class cbq 1:6 parent 1: leaf 6: rate 30000bit (bounded) prio 5
Sent 0 bytes 0 pkts (dropped 0, overlimits 0)
borrowed 0 overactions 0 avgidle 6.10738e+06 undertime 0
class tbf 6:1 parent 6

Izvinite za neredovno odgovaranje na postove, nisam nezainteresovan vec me veza sa internetom veoma mnogo zeza.

 

[ravni]

ok sad se razjasnilo.
kucaj jos
tc -s filter ls dev eth0

pa da vidimo sta treba raditi

 

[koks]

ok sad se razjasnilo.

kako kome meni sav taj tekst malo znaci...

a evo output kada ukucam tc -s filter ls dev eth0

filter parent 1: protocol ip pref 100 u32
filter parent 1: protocol ip pref 100 u32 fh 800: ht divisor 1
filter parent 1: protocol ip pref 100 u32 fh 800::800 order 2048 key ht 800 bkt 0 flowid 1:6
match c0a8011e/ffffffff at 12
match 00000000/00000000 at 16

 

[ravni]

vidi, ni ja se nisam predugo ovim bavio, ali cini mi se da bi u ovome mogao biti problem

match 00000000/00000000 at 16

to bi znacilo da filter preuzima sav odlazni saobracaj sa mreze i salje ga na class 1:6 kome si dodelio 30kbit

ali nesto me jos zbunjuje:
u podacima koje si dao, nigde se ne vidi protok. svuda je 0KB
da li ta mreza radi svo vreme, ili ukljucujes ruter kada je potreban?

 

[ravni]

izgleda da nesto drugo kvari tvoju srecu

http://mailman.ds9a.nl/pipermail/lartc/2002q3/004396.html

>
> The same problem I encountered myself when I try to SNAT some ip's
> going through my router. The incoming is shaped ok but the outgoing
> doesn't seem to hit my rules. I think when the traffic leaves the outgoing
> interface isn't caught in the filter rules because the source address of
> the packets is already changed with the address specified in the SNAT
> rules or with the address of the outgoing interface if you do masquerade.
> To make some shapers, I think you have to shape the ip of the outgoing

Exactly - once tc 'sees' your packets they have everything done to them
already. Every translation by then will have finished.

> interface, but this only works if you want to shape globally only one
> filter rules because there is only one ip of the outgoing interface you can catch it in
> the filter rules. If someone can give any clues how we can solve this
> problem please help us.

Add fwmarks - they survive the NAT process and tell you where packets came
from. The HOWTO details this.

 

[ravni]

tako da bi tvoj problem mozda resilo nesto ovako

iptables -t mangle -A PREROUTING -i eth1 -s 192.168.1.30 -j MARK --set-mark 30

tc filter add dev eth0 protocol ip parent 1:0 prio 1 handle 30 fw flowid 1:6


inace, info je mahom sa www.lartc.org tako da ako zelis da vise udjes u ovo, pogledaj sta ima tamo

 

[koks]

Cim probam, saljem odgovor!

A ruter radi sve vreme, ne gasi se osim kada Elektrodistribucija odluci drugacije.

match 00000000/00000000 at 16 mislim da potice od predloga korisnika Paradigma kada sam stavio RULE=192.168.1.30/32,0.0.0.0/0 kao pravilo.