[c-nsp] If BGP is running on a circuit, if you ping the other end you get loss. kill the BGP (and thus the traffic..) no more loss.

Ibrahim Abo Zaid ibrahim.abozaid at gmail.com
Wed Apr 30 05:33:08 EDT 2008


Hi Drew

when you shut the peering with a neighbor , the routes was received from
this neighbor withdrawn from both BGP or IP routing tables so traffic will
take other available routes so it is normal that the traffic over the trunks
will drop and hence ICMP traffic will find a room .


but the question here , are these links are congested or not ? is there any
QoS policy priorities IP traffic over ICMP traffic ? is locally originated
ICMP traffic is process switched or CEF ? does IOS has a default policy
lessen ICMP priority for sake of IP/TCP/UDP traffic especially in such high
end gears ?


so you need to see the traffic level before and after peering shutdown and
discuss IOS and QoS issues possibilities .

best regards
--Abo Zaid



On 4/29/08, Drew Weaver <drew.weaver at thenap.com> wrote:
>
>                Hi there, I've seen this a few times in the past and its
> always been chalked up to a line or upstream issue but a couple of times
> I've noticed that if I do a ping ip with say 1000 repeats of size 100 I'll
> hit maybe 60% loss on circuits which have BGP neighbors, but if I shutdown
> the BGP neighbor and repeat the "test" the circuit is "clean". I am trying
> to find a 'definitive' way to determine whether or not the issue is that:
>
>
> A)     When I shutdown BGP the traffic on the line dropped to a level in
> which the circuit or the "device" on the other end could actually handle it.
>
> B)      My Router/Line card could handle sending the ICMP because I shut
> down the BGP session on the circuit.
>
> The circuit sizes have ranged from a POS (622Mbps) and a Gig-E So two
> different types of line cards (this is a GSR) so I did a 'show ip cef
> resources' and its all 'G' so I assume that means I am not pushing the line
> cards too hard.
>
> I'm assuming the issue is A but I'd like a way to really know for certain,
> any thoughts?
>
> -Drew
>
>
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/
>


More information about the cisco-nsp mailing list