[LARTC] Resetting traffic history

[LARTC] Resetting traffic history

[Mike Slinn]

[-- Attachment #1: Type: text/plain, Size: 1387 bytes --]

I'm a tc newbie, and I think I am close to being able to use it to 
control one of the virtual web sites on our Gentoo Linux server.  The 
site has it's own IP address.  I have a bit of a problem in that the way 
I originally configured tc, the busy site grabbed all the bandwidth, 
leaving none for the other (and more important) sites.  Here is how I 
had configured it:

    tc qdisc replace dev $NIC root tbf rate $RATE_TOTAL latency 50ms
    burst $BURST

The total data rate was pegged within acceptable limits, but the problem 
is that data stopped flowing after tc was active after a few hours.  The 
busy site had a few peak periods and presumably used up all the traffic 
allotment.  Perhaps tc remembers the traffic between invocations?

I then tried a slightly more sophisticated setup:

    tc qdisc del dev $DEV root
    tc qdisc  add dev $NIC root handle 1: cbq avpkt 1000 bandwidth 1000mbit
    tc class  add dev $NIC parent 1: classid 1:1 cbq rate $RATE_PROBLEM
    allot 1500 prio 5 bounded # isolated
    tc filter add dev $NIC parent 1: protocol ip prio 16 u32 match ip
    dst $IP flowid 1:1

Unfortunately, I still don't get any traffic flowing while tc is active 
now.  Seems that I need to reset something.  Any suggestions?  I've shut 
down the problem site and disabled tc while I try to figure out a solution.

Thanks for your help!

Mike
mslinn at mslinn.com

[-- Attachment #2: Type: text/html, Size: 1876 bytes --]

Re: [LARTC] Resetting traffic history

[Andy Furniss]

Mike Slinn wrote:
> I'm a tc newbie, and I think I am close to being able to use it to 
> control one of the virtual web sites on our Gentoo Linux server.  The 
> site has it's own IP address.  I have a bit of a problem in that the way 
> I originally configured tc, the busy site grabbed all the bandwidth, 
> leaving none for the other (and more important) sites.  Here is how I 
> had configured it:
> 
>    tc qdisc replace dev $NIC root tbf rate $RATE_TOTAL latency 50ms
>    burst $BURST
> 
> The total data rate was pegged within acceptable limits, but the problem 
> is that data stopped flowing after tc was active after a few hours.  The 
> busy site had a few peak periods and presumably used up all the traffic 
> allotment.  Perhaps tc remembers the traffic between invocations?
> 
> I then tried a slightly more sophisticated setup:
> 
>    tc qdisc del dev $DEV root
>    tc qdisc  add dev $NIC root handle 1: cbq avpkt 1000 bandwidth 1000mbit
>    tc class  add dev $NIC parent 1: classid 1:1 cbq rate $RATE_PROBLEM
>    allot 1500 prio 5 bounded # isolated
>    tc filter add dev $NIC parent 1: protocol ip prio 16 u32 match ip
>    dst $IP flowid 1:1
> 
> Unfortunately, I still don't get any traffic flowing while tc is active 
> now.  Seems that I need to reset something.  Any suggestions?  I've shut 
> down the problem site and disabled tc while I try to figure out a solution.
> 
> Thanks for your help!
> 
> Mike
> mslinn at mslinn.com
> 

 From your other post I see 2.6.8.3 - I had to patch 2.6.8.1 to fix a TC 
options related panic. I don't know if it's in 2.6.8.3 already, though.

http://www.linuxhq.com/kernel/v2.6/9-rc2/net/sched/sch_api.c

Andy.


_______________________________________________
LARTC mailing list / LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/

[LARTC] Resetting traffic history

[Michael Slinn]

[-- Attachment #1: Type: text/plain, Size: 1387 bytes --]

I'm a tc newbie, and I think I am close to being able to use it to 
control one of the virtual web sites on our Gentoo Linux server.  The 
site has it's own IP address.  I have a bit of a problem in that the way 
I originally configured tc, the busy site grabbed all the bandwidth, 
leaving none for the other (and more important) sites.  Here is how I 
had configured it:

    tc qdisc replace dev $NIC root tbf rate $RATE_TOTAL latency 50ms
    burst $BURST

The total data rate was pegged within acceptable limits, but the problem 
is that data stopped flowing after tc was active after a few hours.  The 
busy site had a few peak periods and presumably used up all the traffic 
allotment.  Perhaps tc remembers the traffic between invocations?

I then tried a slightly more sophisticated setup:

    tc qdisc del dev $DEV root
    tc qdisc  add dev $NIC root handle 1: cbq avpkt 1000 bandwidth 1000mbit
    tc class  add dev $NIC parent 1: classid 1:1 cbq rate $RATE_PROBLEM
    allot 1500 prio 5 bounded # isolated
    tc filter add dev $NIC parent 1: protocol ip prio 16 u32 match ip
    dst $IP flowid 1:1

Unfortunately, I still don't get any traffic flowing while tc is active 
now.  Seems that I need to reset something.  Any suggestions?  I've shut 
down the problem site and disabled tc while I try to figure out a solution.

Thanks for your help!

Mike
mslinn at mslinn.com

[-- Attachment #2: Type: text/html, Size: 2052 bytes --]