[c-nsp] TCP Throughput / MTU problem ? with Cisco 7304 and MPLS VPN's

[Eric Kagan]

I want to thank everyone for their input and give a quick update.  I created
a better test envorinment that I had full control over (vs being dependent
on a customer setup) and I was able to reproduce the problem.  I ended up
increasing the physical interface MTU (to 1520) vs the MPLS MTU Command
(which I had tried at 1546 and 1520). Once I changed the interface MTU the
traffic passed properly.  I know much of this was mentioned and it was great
to receive such revelant positive feedback to the problem.
 
I then got an update from Cisco TAC that "certain" Gigabit interfaces had a
"problem" and ingored the MPLS MTU command on the 7304 platform.   There is
a documented Bug (below) associated with this issue, which I had tried to
lookup orignally but did not find one.

Thanks again for all the help and info - 
 
Eric
 
 
It looks like you are hitting the internal bug:

CSCeg59156


Increase interface default MTU to 1536
Description: Modified 041217 by bsou
Click to see the enclosure with line wrapping enabled


Problem:
----------
On 7304 traffic with packet size greater than the "interface mtu" gets
dropped when "protocol mtu" is set larger than the "interface mtu", .

This is because the packets are being fragmented based on the protocol
mtu, rather than the interface mtu and the driver code performs sanity
check on the packet size based on the interface mtu.
This results in the following error message being printed.
    NSE100-3-TOOBIG: attempt to send giant packet on GigabitEthernet0/1

So by changing the interface mtu to be the same or greater than the protocol
mtu, the size of the fragmented packet will be equal or less than the
interface
mtu. Thus the packet will go through.


Kind regards,
 

> -----Original Message-----
> From: cisco-nsp-bounces at puck.nether.net 
> [mailto:cisco-nsp-bounces at puck.nether.net] On Behalf Of 
> Mikael Abrahamsson
> Sent: Saturday, January 12, 2008 7:57 PM
> To: cisco-nsp at puck.nether.net
> Subject: Re: [c-nsp] TCP Throughput / MTU problem ? with 
> Cisco 7304 and MPLS VPN's
> 
> On Sat, 12 Jan 2008, Peter Rathlev wrote:
> 
> > As ?ukasz writes, they're changing the commands in newer 
> IOS releases. 
> > That's one argument for doing it.
> 
> I do hope cisco has commands on all platforms to augment all other 
> protocol MTUs so one can emulate the "mtu 1500 / mpls mtu 
> 1546" behaviour 
> that is currently available.
> 
> The situation with devices on a subnet / L2 broadcast domain 
> where some 
> speak mpls and some not, where some are older devices that do 
> not support 
> mtu higher than 1500, needs this behaviour to avoid a network 
> redesign.
> 
> -- 
> Mikael Abrahamsson    email: swmike at swm.pp.se
> _______________________________________________
> 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/
>