Skip to contentSkip to conversation

HTB with SFQ in leaf classes. & setup question

LARTC [email protected] 2 participants 4 messages Dec 10, 2002

lore
Share & tools
lore
#0 lore raw

Stef Coene <[email protected]>:

Quoted 10 lines · On Tuesday 10 December 2002 19:48, [email protected] wrote:
Quoted 8 lines

Hello,

I wrote something yesterday.
Now the situation is even worse. Communicatin with some clients wont work at all until i restart the traffic shaping box.
(It works for some hours, then suddenly dies (meaning the client cant be pinged. Suspected at peak hours)
Maybe the computer resources are weak, maybe its setup faulty

Do you have any errors in dmesg? And how is the memory and cpu usage ?

Right now (night) the traffic shaping box is not loaded at all 10% cpu usage and 200MB ram used.
Tho at the daytime cpu usage goes up to 70% or so. and ~ 500MB ram.

Yes there are errors.

Dec 10 22:59:28 p1 kernel: NET: 26 messages suppressed.
Dec 10 22:59:28 p1 kernel: HTB: bad diff in events, cl243 diffE26000 nowC4192384 

thenB94893568 j01578
...

Like this.

Just before i left work, i left a file d/ling, to test what happens.
Right now i cant reach outside world from that computer (128MB with speed of 128kbit/s have downloaded and it has halted).

I can reach the computer from the computer where the traffic shaping is being done. But not from a computer which is a hop from traffic shaping computer

internet - router1 - traffic_shaping - ... - me

80.81.33.1 is the router1 at the ISP providing us 8mbit like. It cant be reached from the box which was being shaped.

me@myplace$ ping 80.81.33.1
PING 80.81.33.1 (80.81.33.1): 56 octets data

--- 80.81.33.1 ping statistics ---
3 packets transmitted, 0 packets received, 100% packet loss

Tho it can be reached from box next to me. (my other workstation)

me@myotherpc$ ping 80.81.33.1 
PING 80.81.33.1 (80.81.33.1): 56 octets data
64 octets from 80.81.33.1: icmp_seq=0 ttl%2 timeb.6 ms
64 octets from 80.81.33.1: icmp_seq=1 ttl%2 time .8 ms

--- 80.81.33.1 ping statistics ---
2 packets transmitted, 2 packets received, 0% packet loss round-trip min/avg/max = 20.8/41.7/62.6 ms

So there is a problem in the traffic shaping box.
Traceroute also stops at the traffic shaping box.

Huge problem.

Ok, just a second ago i did '/sbin/tc qdisc del dev eth2 root' And added the classes again.

me@myplace$ ping 80.81.33.1
PING 80.81.33.1 (80.81.33.1): 56 octets data
64 octets from 80.81.33.1: icmp_seq=0 ttl%2 time$.6 ms
64 octets from 80.81.33.1: icmp_seq=1 ttl%2 time!.9 ms

--- 80.81.33.1 ping statistics ---
2 packets transmitted, 2 packets received, 0% packet loss round-trip min/avg/max = 21.9/23.2/24.6 ms

I can reach ISP's gateway very smoothly now!

Any idee?

P.Krumins
____________________________________________________________

Reìistrç un lieto savu Apollo e-pastu www.e-apollo.lv

www.apollo.lv

_______________________________________________
LARTC mailing list / [email protected]
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/

#1 lore raw
Quoted 8 lines · On Tue, 10 Dec 2002 23:27:39 +0200 [email protected] wrote:

Ok, just a second ago i did '/sbin/tc qdisc del dev eth2 root' And added the classes again.

me@myplace$ ping 80.81.33.1
PING 80.81.33.1 (80.81.33.1): 56 octets data
64 octets from 80.81.33.1: icmp_seq=0 ttl=252 time=24.6 ms
64 octets from 80.81.33.1: icmp_seq=1 ttl=252 time=21.9 ms
Quoted 5 lines

--- 80.81.33.1 ping statistics ---
2 packets transmitted, 2 packets received, 0% packet loss round-trip min/avg/max = 21.9/23.2/24.6 ms

I can reach ISP's gateway very smoothly now!

Hello, I'm running a similar setup with about a thousand clients, separated in several level
hierarchy. The machine is also similar. I had the same problem half a year ago.

Anyway, can you tell us what is the version of htb? It's htb2, isn't it? Try upgrading to htb3, this will solve your problem. If not, just put a crontab job to restart (i.e. delete and the recreate) the htb classes every 1/2 hour.

Hth,
Nickola

#2 lore raw
In reply to [email protected]

Nickola Kolev <[email protected]>:

Quoted 17 lines
Quoted 8 lines · On Tue, 10 Dec 2002 23:27:39 +0200 [email protected] wrote:

Ok, just a second ago i did '/sbin/tc qdisc del dev eth2 root' And added the classes again.

me@myplace$ ping 80.81.33.1
PING 80.81.33.1 (80.81.33.1): 56 octets data
64 octets from 80.81.33.1: icmp_seq=0 ttl%2 time$.6 ms
64 octets from 80.81.33.1: icmp_seq=1 ttl%2 time!.9 ms
Quoted 5 lines

--- 80.81.33.1 ping statistics ---
2 packets transmitted, 2 packets received, 0% packet loss round-trip min/avg/max = 21.9/23.2/24.6 ms

I can reach ISP's gateway very smoothly now!

Quoted 8 lines

Hello, I'm running a similar setup with about a thousand clients, separated in several level
hierarchy. The machine is also similar. I had the same problem half a year
ago.

Anyway, can you tell us what is the version of htb? It's htb2, isn't it? Try upgrading to htb3, this
will solve your problem.

No, it's htb3.6_2.4.17.diff
HTB 3.6
Tho it's made for 2.4.17 kernels, i am running 2.4.19.

If not, just put a crontab job to restart (i.e.

Quoted 2 lines

delete and the recreate)
the htb classes every 1/2 hour.

Yes, i just did it. I hope 1/2 hour is reasonable interval. Will see what the clients gonna do tomorro.

Btw, how do you solve this problem?

I am about to test it more with 2.4.20 (HTB included) on Sunday

P.Krumins
____________________________________________________________

Reìistrç un lieto savu Apollo e-pastu www.e-apollo.lv

www.apollo.lv

_______________________________________________
LARTC mailing list / [email protected]
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/

#3 lore raw
In reply to [email protected]

On Tue, 10 Dec 2002 23:58:03 +0200
[email protected] wrote:

[cut]

Quoted 7 lines
Quoted 3 lines

Anyway, can you tell us what is the version of htb? It's htb2, isn't it? Try upgrading to htb3, this
will solve your problem.

No, it's htb3.6_2.4.17.diff
HTB 3.6
Tho it's made for 2.4.17 kernels, i am running 2.4.19.

[cut]

Hmpf, strange. I'm really sorry that I didn't test and debug this when it was happening, but this is a production machine and I had to find a solution fast. But now I'm rid of it (after upgrading to htb3 in may).

[cut]

Quoted 9 lines
Quoted 2 lines

delete and the recreate)
the htb classes every 1/2 hour.

Yes, i just did it. I hope 1/2 hour is reasonable interval. Will see what the clients gonna do tomorro.

Btw, how do you solve this problem?

I am about to test it more with 2.4.20 (HTB included) on Sunday

[cut]

Don't worry, half an hour is quite reasonable. :)

And about the solution... emmm... I just upgraded, that's all.

Nickola