[Netem] Re-order packets problems
Stephen Hemminger
shemminger at osdl.org
Wed Feb 8 11:07:18 PST 2006
On Wed, 8 Feb 2006 09:51:31 -0000
"Maria de Lurdes Sousa" <Lurdes.Sousa at siemens.com> wrote:
> Hi *,
> i´m trying to some tests with my network and for this i need to send some out-of-order packets!
> To emulate this i use the following command:
>
> #tc qdisc change dev eth0 root netem gap 2 delay 300ms
>
> when i try to make a ping i see that all packets have more 300ms, not just the second.
> Someone can help me about this, please?
>
> And another question how can i reload the qdisc?
>
> I have this version of FC4
> # uname -a
> Linux alcoholix 2.6.15-1.1830_FC4 #1 Thu Feb 2 17:23:41 EST 2006 i686 i686 i386 GNU/Linux
>
> Thanks in advance for your support
> Maria
>
>
The "gap" was really enhanced by a more general function based on reordering
probability in later kernels.
# tc qdisc add dev eth0 root netem reorder 100% gap 5 delay 100ms
# ping shell
64 bytes from shell0.pdx.osdl.net (10.9.0.31): icmp_seq=0 ttl=63 time=101 ms
64 bytes from shell0.pdx.osdl.net (10.9.0.31): icmp_seq=1 ttl=63 time=101 ms
64 bytes from shell0.pdx.osdl.net (10.9.0.31): icmp_seq=2 ttl=63 time=1.11 ms
64 bytes from shell0.pdx.osdl.net (10.9.0.31): icmp_seq=3 ttl=63 time=101 ms
64 bytes from shell0.pdx.osdl.net (10.9.0.31): icmp_seq=4 ttl=63 time=101 ms
64 bytes from shell0.pdx.osdl.net (10.9.0.31): icmp_seq=5 ttl=63 time=101 ms
64 bytes from shell0.pdx.osdl.net (10.9.0.31): icmp_seq=6 ttl=63 time=101 ms
64 bytes from shell0.pdx.osdl.net (10.9.0.31): icmp_seq=7 ttl=63 time=101 ms
64 bytes from shell0.pdx.osdl.net (10.9.0.31): icmp_seq=8 ttl=63 time=1.12 ms
64 bytes from shell0.pdx.osdl.net (10.9.0.31): icmp_seq=9 ttl=63 time=101 ms
You may need to rebuild iproute2 utils to get at it. Also, watch out
for the fact that iproute2 utils "make install" puts the new tc
in /usr/sbin/tc, but FC4 ships with /sbin/tc so you may get the old
version.
--
Stephen Hemminger <shemminger at osdl.org>
OSDL http://developer.osdl.org/~shemminger
More information about the Netem
mailing list