[c-nsp] Unexplainable packet loss

[Pavel Skovajsa]

On Sun, Sep 19, 2010 at 2:36 AM, ML <ml at kenweb.org> wrote:
>  On 9/18/2010 6:28 AM, Heath Jones wrote:
>>
>> Hi
>> Firstly, when you say packet loss, what are you referring to? Is it just
>> the icmp traffic, or are customers reporting faults with non icmp traffic
>> or...?
>> Is the 'internet gateway' the 7609 pictured on the diagram?
>> Its pretty unlikely, but worth checking that there are no duplicate mac
>> addresses on the network. A duplicate (of 7609 on mdf side) could cause
>> these symptoms.
>> You could swap out the RAD with your own device for testing..
>> I don't think that standard icmp tests will identify the problem though.
>> If what the engineer said is true, then you really need to be pushing some
>> traffic down to see it. (load related issue)
>> 'Another engineer tells me that when our customers traffic is removed from
>> the picture the packet loss goes away'
>> The first thing though - what is the packet loss?
>
> The customer is reporting a problem. They have an outside IT service that
> monitors a firewall/VPN solution for them.
> We never went into detail about what kind of packetloss they are seeing
> since the problem appears to be on our side/our upstream.
>
> Yes the 'internet gateway' is the 7609.  The 7609 is the device with the L3
> interface we use as a default route.
> ICMP packetloss anywhere from 1-5% when a set of 1000 pings are sent from
> MDF to 7609 L3 interface.

What happens when they stop pinging your 7609 and start pinging their
own device (on the other side) of the link? It can easily happen that
somebody else is pinging the 7609 too, resoluting in some CPU MLS
rate-limiter (show mls rate-limit, show mls rate-limit usage) kicking
into action, dropping your ICMP reply packets.

-pavel

> _______________________________________________
> 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/
>