Folks, I'm so surprised what happened to my box just in the few weeks lately. Here is my setup:
INTERNET <----------> [eth0] SHAPER-BOX [eth1] <----------> USER-FARM both eth0 & eth1 got public ips (202.x.x.x)
Why traffic monitored at eth0 is bigger than eth1 ? eth1 shaped just exactly the same as rate i defined in HTB.
I just have 1024Kbps from my ISP and i defined the total parent rate for 768Kbps. But what happened is the traffic monitored at eth0 almost all the time eating all the bandwidth we have.
# DOWNSTREAM 768Kbps
/sbin/ip link set imq1 up
/sbin/tc qdisc add dev imq1 root handle 2: htb default 0
/sbin/tc class add dev imq1 parent 2: classid 2:1 htb rate 768Kbit
/sbin/tc class add dev imq1 parent 2:1 classid 2:10 htb rate 128kbit ceil 768kbit
/sbin/tc class add dev imq1 parent 2:1 classid 2:20 htb rate 128kbit ceil 768kbit
/sbin/tc filter add dev imq1 parent 2: protocol ip handle 10 fw classid 2:10
/sbin/tc filter add dev imq1 parent 2: protocol ip handle 20 fw classid 2:20
/usr/sbin/iptables -t mangle -A PREROUTING -i eth0 -j IMQ --todev 1
/usr/sbin/iptables -t mangle -A PREROUTING -i eth0 -j IMQ --todev 1
You only need one -j IMQ here.
/usr/sbin/iptables -t mangle -A PREROUTING -i eth0 -d 202.x.1.0/24 -j MARK --set-mark 10
/usr/sbin/iptables -t mangle -A PREROUTING -i eth0 -d 202.x.2.0/24 -j MARK --set-mark 20
You will get away with the mark after -j imq - but it doesn't look very elegant :-)
Please help me analyze whats wrong.
Assuming tc -s class or qdisc ls dev imq1 shows everything is getting there OK then I think the problem is that htbs quelength is far too long for your link.
You are also shaping from the wrong end of the bottleneck so the queues only fill slowly.
Add queues to your htb leaf classes so you can set the length, if you choose sfq use the limit option as the 128 packet default is too long.
Andy.
_______________________________________________ LARTC mailing list LARTC@xxxxxxxxxxxxxxx http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
