Hi przemolicc@2004.07.12_09:18:07_+0200
On a side note has anyone built a linux router with dual/quad port
ethernet cards (ie Intel PRO/1000 MT Quad Port Server Adapter)?
I have bulit a linux router with quad D-Link DFE-580TX nics.
It works like a
Hi!
How does prioritization work when you've got a tree structure, e.g.
|
+-- class_a rate 64kbit prio 1
||
|+-- class_a1 rate 32kbit prio 1
||
|`-- class_a2 rate 32kbit prio 2
|
`-- class_b rate 64kbit prio 2
Above could either be interpreted as
(a) a, a1 have prio 1
bandwidth being much larger than your interface actually is and then
subdividing into your groups?
Mike.
-Original Message-
From: Abraham van der Merwe [mailto:[EMAIL PROTECTED]
Sent: Monday, May 31, 2004 8:58 AM
To: Linux Advanced Routing Traffic Control
Subject: [LARTC
Hi!
Is there any way to do just plain vanilla TBF (Token Buck Filter) type
shaping on a group of ips/networks, not an entire interface.
Currently the only way I know how to shape in Linux is to use HTB or CBQ,
but both of these need a total rate and then you need to subdivide that
into classes.
Hi Mahmoud @2003.09.22_11:44:01_+0200
(a) You can't put it in PREROUTING - the CLASSIFY patch don't allow that:
from net/ipv4/netfilter/ipt_CLASSIFY.c:
snip -- snip -- snip
if (hook_mask ~(1 NF_IP_POST_ROUTING)) {
Hi!
This morning one of our major links died and suddenly I was streaming 1mbit
of traffic through 256kbit (one of our backup links), but it just doesn't
work. The client is still doing in the region of 350kbit/s (the link
capacity is 512kbit). Here is my config:
snip -- snip
Hi hare!
tc qdisc add dev eth0 root handle 1: htb default 2 r2q 10
tc class add dev eth0 parent 1: classid 1:1 htb rate 256kbps
tc class add dev eth0 parent 1:1 classid 1:2 htb rate 64kbps ceil 64kbps
tc qdisc add dev eth0 handle 2: parent 1:2 sfq perturb 10 limit 178
tc class add dev eth0 parent
Hi!
I have the following setup:
ISP 1 ISP 2
+-- +--
\ /
\ /
+--+
| router A |
+--+
/\
\
Hi Eric!
iptables -A FORWARD -s $net -m conntrack --proto ftp
iptables -A FORWARD -s $net -m conntrack --proto irc
iptables -A FORWARD -s $net -m conntrack --proto h323
To do so you can use the conmarck module (from iptable pom) : the mark
of the packet is given following the conntrack.
Hi!
Usually if you have a machine and traffic passes through it:
+-+
eth0| QoS |
-| box |-
| |eth1
+-+
You can shape outgoing traffic on eth0 and eth1 effectively shaping both
incoming/outgoing traffic.
is
used only for trafic that is for the bridge. Eg. a machine has a route
through the ip of br0.
Bogdan Coman
On Tue, 2003-03-04 at 10:58, Abraham van der Merwe wrote:
Hi!
Usually if you have a machine and traffic passes through it:
+-+
eth0| QoS
Hi S!
The packet flow diagrams have been well documented here.
http://users.pandora.be/bart.de.schuymer/ebtables/br_fw_ia/br_fw_ia.html
That assumes you're using ebtables and matching traffic passing through br0.
I want to match traffic passing through eth0 and eth1.
If you are matching
Hi!
If I create the following setup:
66.8.28.52/2966.8.28.51/29
+--+ +--+
| PC A |--+ +-| PC B |
+--+ | | +--+
| |
eth1| | eth0
+-+
| qos | (br0 =
Hi Martin!
: No, I'm not running with ebtables+nf support. From what I understand
: (and please correct me if I'm wrong), patching the kernel with
: ebtables+bridge-nf, you get an ebtables table with BROUTING, FORWARD,
: and NAT chains which you can match traffic on.
:
: However, I
Hi Martin!
I just applied the bridge-nf and ebtables patches and tried it and I can
match packets in the mangle table as usual (also have to use FORWARD for
packets passing through the machine).
bill-the-cat-sound Ack! I meant to say:
It sounds like you are running bridging without the
the machines are physically located far from here.
I realize that this might not be enough information to make something useful of.
It is merely an attempt to somewhat document the problem.
Göran
- Original Message -
From: Abraham van der Merwe [EMAIL PROTECTED]
To: Göran Runfeldt
Hi Nedko!
i have problem then i use classid bigger that .
then i make htb classid 1:1 i see that in syslog class htb*c10001
then i make htb classid 1:10001 i get error class exist ???
looks like classid 1:1 and classid 1:10001 create same class
is this normal ?
a) the classid's are in
Hi!
I've got a box which redirects http traffic to another box (proxy). I'd like
to do some traffic shaping on this box as well (including traffic destined
for the proxy). In order to do that I need to intercept packets in the
mangle table in PREROUTING, shape it and then it will be redirected
Hi Patrick!
ok this one actually compiles ;)
just a small bug in your code:
+ *p = TC_H_MAKE(i, j);
that should be changed to TC_H_MAKE(i16, j)
--
Regards
Abraham
Here we are in America ... when do we collect unemployment?
___
Hi Patrick!
ok this one actually compiles ;)
just a small bug in your code:
+ *p = TC_H_MAKE(i, j);
that should be changed to TC_H_MAKE(i16, j)
thanks, i've attached a changed diff. have you actually tested it ?
does it work ? ;)
Yes, I tested it last
Hi Thomas!
The smallest rate you can use with r2q=1 is mtu*8 kbit (12kbit in most
cases). You can skip r2q and use quantum=125 bytes, but that is probably not
such a good idea...
OK, I will use rate 1kbit. Therefore r2q should be 1, right?
Thanks,
Thomas
On Fri, 10
Hi!
A couple of weeks ago we started using the Linux TC implementation to shape
clients (using HTB) and since then our QoS box started crashing every now
and again (talk about inconvenient timing :P).
First we suspected hardware, but we replaced the hardware this week and it
still keeps
Hi Stef!
I did some small tests :
[snip]
Conclusion : if you want to know how traffic will be shaped, you have to
follow some basic rules.
I will update docum.org tonight.
You're right. I did some tests this weekend as well and realised that you
have to stick with the basics. The problem
Hi Stef!
How does this influence the shaping results and the shaping of other
classes? I've actually asked a similar question before: If you have
1:1
/ \
/ \
1:2 1:3
/|\ \--
/ | \ \ \
1:4
Hi Don!
I then tried fifos. With small packet fifos the packet loss is just
to great to be of any use and even then the latency is quite high (~200ms).
You consider 200ms high? One max size packet = 1500 bytes = 12Kbit
which is about 200ms on a 64Kbit link. You can't expect to do
Hi!
I'm using HTB to shape traffic to/from clients, but one of the problems I
have is that once a class utilizes its maximum potential, they latency
spirals out of control.
For example:
.-.
| isp |
`-'
|
..
| shaper |
`'
|
Hi Andrei!
Look in the mail archives. Somebody posted a solution for GRE tunnels last
week.
After carefull reading (LARTC) and experimentation, I am in a dead
end...
I am using several IPIP tunnels (linux ipip module, IP protocol 4).
I'd like to filter packets going through these
Hi Roch?!
I just had a data line installed between my home and the office. I've
managed to connect the two sides successfully with ppp and the gateways
on both sides can see each other. I want the hosts behind the
gateways to see each other as well and although this is discussed in
some
Hi!
HTB questions:
1. why is quantum not always: quantum=mtu?
From http://luxik.cdi.cz/~devik/qos/htb/manual/userg.htm:
snip -- snip -- snip
It is important to know that for precise operation quantums need to be as
small as possible and larger than MTU
Hi Stef!
HTB questions:
1. why is quantum not always: quantum=mtu?
From http://luxik.cdi.cz/~devik/qos/htb/manual/userg.htm:
snip -- snip -- snip
It is important to know that for precise operation quantums need to be as
small as possible and
Hi Stef!
Hmm, lets say I want to do this:
1. shape/prioritize some subnets according to some rules
2. shape/prioritize some protocols according to some rules
(1) should be evaluated and then the data stream should be passed on to (2)
and be evaluated again according to that set of
Hi!
I have a machine
ISP
|
| eth0
+---+
| Linux | eth2
| QoS |-- DMZ
| box |
+---+
| eth1
|
WAN
I want to shaping incoming bandwidth and outgoing bandwidth where incoming
means all packets entering eth0 and outgoing means all packets entering eth1
_or_
Hi!
I want to route packets where the gateway is determined by the source
address and not the destination address - is this possible with iproute2?
for example:
ISP-A ISP-B
||
+---+
| linux machine +
+---+
default gw is ISP-A interface, but if packet comes
Hi Juergen!
I just tried to upgrade to linux 2.4.20-rc2 and found, that my scripts are
not working any more (they did with 2.4.18 + HTB patch).
It boils down to:
# tc qdisc add dev ppp0 root handle 1: htb default 12
RTNETLINK answers: Invalid argument
while
# tc qdisc add dev ppp0
Hi David!
I prefer using shaping instead of policing in most cases (in my experience
it is always a hassle tuning rate-limit's to handle tcp streams correctly).
For shaping you can use HTB (just remember that it can only delay outgoing
traffic, so you need to add rules on eth1 in your case for
Hi Juergen!
I got HTB working with 2.4.20-rc1 yesterday. Make sure you're using the
latest iproute with HTB3 patches applied. (ss020116 + htb3 patches iirc)
Do I need those patches with the HTB included in the kernel?
Is HTB 2 or 3 in 2.4.20 ?
No, not if you use 2.4.20rcXX. It already
Hi!
I'm trying to use tc on linux 2.4.20-rc1aa with iproute 20010824-8 (latest
debian package), but I'm having no joy.
I can add queueing disciplines, but when I try to add classes or filters, I
just get RTNETLINK answers: Invalid argument, e.g. when I try to do:
snip -- snip
Hi Stef!
I'm trying to shape a traffic to/from a few different subnets using TBF.
The way I understand it all I have to do is create a few queueing
disciplines and attach a filter to each queueing discipline.
So for instance if I want to shape traffic going to 192.168.1.0/24 via eth0
38 matches
Mail list logo
