Wonder Shaper problem?
I think I have an issue with the Wonder Shaper script. I setup my firewall to use the Wonder Shaper script with HTB (instead of CBQ). My firewall has Squid running in transparent proxy mode, but I setup my browser to point to port 3128 (default Squid port). I ran a ping session to www.yahoo.com and did an SSH at the same time that I downloaded the Linux kernel (~20MB). It appears that most (if not all) of the traffic is being placed into the high priority class (queue? don't really know the difference yet) 10:. Isn't it supposed to go to 20: if it's not SSH or ICMP or ACK packets? Here are the ping times from Yahoo! while doing the kernel download:
sh-2.05# ping www.yahoo.com
PING www.yahoo.akadns.net (64.58.76.227): 56 data bytes 64 bytes from 64.58.76.227: icmp_seq=0 ttl$1 time59.8 ms 64 bytes from 64.58.76.227: icmp_seq=1 ttl$1 time93.4 ms 64 bytes from 64.58.76.227: icmp_seq=2 ttl$1 time 18.8 ms 64 bytes from 64.58.76.227: icmp_seq=3 ttl$1 time72.8 ms 64 bytes from 64.58.76.227: icmp_seq=4 ttl$1 time95.7 ms 64 bytes from 64.58.76.227: icmp_seq=5 ttl$1 time 03.9 ms 64 bytes from 64.58.76.227: icmp_seq=6 ttl$1 time 26.3 ms 64 bytes from 64.58.76.227: icmp_seq=7 ttl$1 time 49.0 ms 64 bytes from 64.58.76.227: icmp_seq=8 ttl$1 time 75.7 ms
--- www.yahoo.akadns.net ping statistics ---
11 packets transmitted, 9 packets received, 18% packet loss round-trip min/avg/max = 1872.8/1988.3/2075.7 ms
As you can see, they are very high, on the order of 2 seconds!!!??
When I stopped the kernel download, the ping results dropped to:
sh-2.05# ping www.yahoo.com
PING www.yahoo.akadns.net (64.58.76.176): 56 data bytes 64 bytes from 64.58.76.176: icmp_seq=0 ttl$1 time8.1 ms 64 bytes from 64.58.76.176: icmp_seq=1 ttl$1 time9.1 ms 64 bytes from 64.58.76.176: icmp_seq=2 ttl$1 time9.1 ms 64 bytes from 64.58.76.176: icmp_seq=3 ttl$1 time9.3 ms 64 bytes from 64.58.76.176: icmp_seq=4 ttl$1 time9.6 ms 64 bytes from 64.58.76.176: icmp_seq=5 ttl$1 time9.2 ms
--- www.yahoo.akadns.net ping statistics ---
6 packets transmitted, 6 packets received, 0% packet loss round-trip min/avg/max = 178.1/179.0/179.6 ms
Here is the output of 'tc -s qdisc list':
sh-2.05# tc -s qdisc list
qdisc ingress ffff: dev eth0
Sent 0 bytes 0 pkts (dropped 0, overlimits 0)
qdisc sfq 20: dev eth0 quantum 1514b perturb 10sec Sent 44775 bytes 401 pkts (dropped 0, overlimits 0)
qdisc sfq 10: dev eth0 quantum 1514b perturb 10sec Sent 166751 bytes 2494 pkts (dropped 0, overlimits 0)
qdisc htb 1: dev eth0 r2q 10 default 20 dcache 0 deq_util 1/1000000 deq_rate 0 trials_per_deq 0 dcache_hits 0 direct_packets 0
Sent 211526 bytes 2895 pkts (dropped 0, overlimits 0)
During the download and the ping, the SSH session is extremely sluggish.
But once both are stopped, the SSH session is very usable.
I'm using the 2.4.16 kernel patched with the HTB patch.
Signature · 8 lines
Jason A. Pattie [email protected] _______________________________________________ LARTC mailing list / [email protected] http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://ds9a.nl/lartc/
Quoted 10 lines · Jason A. Pattie
I think I have an issue with the Wonder Shaper script. I setup my firewall to use the Wonder Shaper script with HTB (instead of CBQ). My firewall has Squid running in transparent proxy mode, but I setup my browser to point to port 3128 (default Squid port). I ran a ping session to www.yahoo.com and did an SSH at the same time that I downloaded the Linux kernel (~20MB). It appears that most (if not all) of the traffic is being placed into the high priority class (queue? don't really know the difference yet) 10:. Isn't it supposed to go to 20: if it's not SSH or ICMP or ACK packets? Here are the ping times from Yahoo! while doing the kernel download:
tcpdump a bit - it may be that squid is misbehaving and giving its traffic 'minimum delay' TOS!
tcpdump -n -v -v
Regards,
bert
Signature · 8 lines
http://www.PowerDNS.com Versatile DNS Software & Services http://www.tk the dot in .tk Netherlabs BV / Rent-a-Nerd.nl - Nerd Available - Linux Advanced Routing & Traffic Control: http://ds9a.nl/lartc _______________________________________________ LARTC mailing list / [email protected] http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://ds9a.nl/lartc/
Would that require tcpdump on the firewall? or could I run tcpdump somewhere else on the network? I'm trying to keep as many tools off the firewall as possible. But I'll put tcpdump into the distro for testing.
Quoted 23 lines · bert hubert
Quoted 11 lines · Jason A. Pattie
I think I have an issue with the Wonder Shaper script. I setup my firewall to use the Wonder Shaper script with HTB (instead of CBQ). My firewall has Squid running in transparent proxy mode, but I setup my browser to point to port 3128 (default Squid port). I ran a ping session to www.yahoo.com and did an SSH at the same time that I downloaded the Linux kernel (~20MB). It appears that most (if not all) of the traffic is being placed into the high priority class (queue? don't really know the difference yet) 10:. Isn't it supposed to go to 20: if it's not SSH or ICMP or ACK packets? Here are the ping times from Yahoo! while doing the kernel download:
tcpdump a bit - it may be that squid is misbehaving and giving its traffic 'minimum delay' TOS!
tcpdump -n -v -v
Regards,
bert
Signature · 9 lines
Jason A. Pattie [email protected] _______________________________________________ LARTC mailing list / [email protected] http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://ds9a.nl/lartc/
Where in the output of tcdump does the TOS field appear? I did the 'tcpdump -n -v -v' on the firewall and dumped the output to a file as I started the Linux kernel download. Is it appropriate to attach the output (I stopped it after around 250 packets)?
Quoted 23 lines · bert hubert
Quoted 11 lines · Jason A. Pattie
I think I have an issue with the Wonder Shaper script. I setup my firewall to use the Wonder Shaper script with HTB (instead of CBQ). My firewall has Squid running in transparent proxy mode, but I setup my browser to point to port 3128 (default Squid port). I ran a ping session to www.yahoo.com and did an SSH at the same time that I downloaded the Linux kernel (~20MB). It appears that most (if not all) of the traffic is being placed into the high priority class (queue? don't really know the difference yet) 10:. Isn't it supposed to go to 20: if it's not SSH or ICMP or ACK packets? Here are the ping times from Yahoo! while doing the kernel download:
tcpdump a bit - it may be that squid is misbehaving and giving its traffic 'minimum delay' TOS!
tcpdump -n -v -v
Regards,
bert
Signature · 9 lines
Jason A. Pattie [email protected] _______________________________________________ LARTC mailing list / [email protected] http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://ds9a.nl/lartc/
Quoted 4 lines · Jason A. Pattie
Where in the output of tcdump does the TOS field appear? I did the 'tcpdump -n -v -v' on the firewall and dumped the output to a file as I started the Linux kernel download. Is it appropriate to attach the output (I stopped it after around 250 packets)?
In the version I have of tcpdump TOS only shows up when its set to something special.
Signature · 7 lines
Michael T. Babcock CTO, FibreSpeed Ltd. (Hosting, Security, Consultation, Database, etc) http://www.fibrespeed.net/~mbabcock/ _______________________________________________ LARTC mailing list / [email protected] http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://ds9a.nl/lartc/
Found the TOS field was being set to 0x10 for "ftp" traffic and 0x08 for "http" traffic. What significance does that play? And it is either being done by Squid or by Mozilla, don't know how to tell which, though.
Quoted 11 lines · Michael T. Babcock
Quoted 5 lines · Jason A. Pattie
Where in the output of tcdump does the TOS field appear? I did the 'tcpdump -n -v -v' on the firewall and dumped the output to a file as I started the Linux kernel download. Is it appropriate to attach the output (I stopped it after around 250 packets)?
In the version I have of tcpdump TOS only shows up when its set to something special.
Signature · 9 lines
Jason A. Pattie [email protected] _______________________________________________ LARTC mailing list / [email protected] http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://ds9a.nl/lartc/
So then is there a way to modify the packet on the way out of squid to force it to be the TOS we want that will cause it to go to 20: instead of 10: ? Maybe modification of squid is in order? to keep it from changing the TOS field? (We have the source! :) )
Or is there some other way to make it so that SSH traffic itself is bumped to higher priority than just the generic 0x10 TOS packets? Maybe match on destination or source port 22?
Quoted 38 lines · Dhaval Patel
i could be wrong its been some time since i did this but i think that
0x10is for quick response time0x08is for high throughput0x04is for a more guranteed service or something like that.hope this helps.
Quoted 27 lines · Jason A. Pattie
Found the TOS field was being set to
0x10for "ftp" traffic and0x08for "http" traffic. What significance does that play? And it is either being done by Squid or by Mozilla, don't know how to tell which, though.Quoted 10 lines · Michael T. Babcock
Quoted 5 lines · Jason A. Pattie
Where in the output of tcdump does the TOS field appear? I did the 'tcpdump -n -v -v' on the firewall and dumped the output to a file as I started the Linux kernel download. Is it appropriate to attach the output (I stopped it after around 250 packets)?
In the version I have of tcpdump TOS only shows up when its set to something special.
Signature · 9 lines
Jason A. Pattie [email protected] _______________________________________________ LARTC mailing list / [email protected] http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://ds9a.nl/lartc/
Signature · 9 lines
Jason A. Pattie [email protected] _______________________________________________ LARTC mailing list / [email protected] http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://ds9a.nl/lartc/
you can have your firewall set the tos for ftp to something like 0x08 and keep only low bandwidth high priority things like ssh as 0x10.
Quoted 61 lines · Jason A. Pattie
So then is there a way to modify the packet on the way out of squid to force it to be the TOS we want that will cause it to go to 20: instead of 10: ? Maybe modification of squid is in order? to keep it from changing the TOS field? (We have the source! :) )
Or is there some other way to make it so that SSH traffic itself is bumped to higher priority than just the generic
0x10TOS packets? Maybe match on destination or source port 22?Quoted 38 lines · Dhaval Patel
i could be wrong its been some time since i did this but i think that
0x10is for quick response time0x08is for high throughput0x04is for a more guranteed service or something like that.hope this helps.
Quoted 27 lines · Jason A. Pattie
Found the TOS field was being set to
0x10for "ftp" traffic and0x08for "http" traffic. What significance does that play? And it is either being done by Squid or by Mozilla, don't know how to tell which, though.Quoted 10 lines · Michael T. Babcock
Quoted 5 lines · Jason A. Pattie
Where in the output of tcdump does the TOS field appear? I did the 'tcpdump -n -v -v' on the firewall and dumped the output to a file as I started the Linux kernel download. Is it appropriate to attach the output (I stopped it after around 250 packets)?
In the version I have of tcpdump TOS only shows up when its set to something special.
Signature · 9 lines
Jason A. Pattie [email protected] _______________________________________________ LARTC mailing list / [email protected] http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://ds9a.nl/lartc/Signature · 9 lines
Jason A. Pattie [email protected] _______________________________________________ LARTC mailing list / [email protected] http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://ds9a.nl/lartc/
Signature · 3 lines
_______________________________________________ LARTC mailing list / [email protected] http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://ds9a.nl/lartc/