[c-nsp] ip tcp adjust-mss

[Eric A Louie]

Apparently, packets traversing the mpls ip interfaces that I have defined are 
getting fragmented.  Cisco hasn't determined "why" yet - I have an open ticket 
on it.

Their solution/workaround was to use this command.  It was not my idea.  (I'm 
not smart enough to even figure out that I could increase MTU on the two mpls ip 
interfaces...)

 Much appreciated, Eric




________________________________
From: Richard Clayton <sledge121 at gmail.com>
To: Eric A Louie <elouie at yahoo.com>
Cc: Cisco NSP <cisco-nsp at puck.nether.net>
Sent: Tue, February 12, 2013 5:04:19 AM
Subject: Re: [c-nsp] ip tcp adjust-mss

Eric

I needed to use this command the other day, I have an 887VA-M and the BT FTTC 
product, I bypassed the BT modem and connected directly into the BT wall socket 
with the 887VA-M as it has a VDSL interface (just a config tweek)

The config I was using was PPPOE which adds 8 bytes to the frame so on my 
dialer0 I set the mtu to 1492, without the 'ip tcp adjust-mss 1452' I could not 
open web pages or run speed tests as my local host's mtu defaulted to their 
local setting (1500) and was dropped.

After adding the command to the LAN interface the router intercepts the SYN 
packet from the local host and changes the maximum segment size to the value 
stated before passing on to the remote host, the TCP 3 way handshake is then 
completed with the two hosts agreeing on the lowest of their 2 values, obviously 
it doesnt help if you have large UDP packets but there shouldn't be too many of 
those around anyway.

Using this command reduces the mtu size for TCP traffic flowing through the 
configured router but in your case I would be more interested in why you think 
you need it and where, do you think you have mtu bottlenecks in your network 
that are causing fragmentation and if so can you just fix those areas rather 
than adding this to lots of other routers.

Thanks
Sledge




On 11 February 2013 19:56, Eric A Louie <elouie at yahoo.com> wrote:

I just put in this command on my upstream interfaces to help my mpls network
>pass traffic - that is, my effort to eliminate fragmentation in my backbone.
>
>Is anyone else using this method of "mtu control"?  I need some support - my 
CEO
>is asking why I have to do this, and who else does it, and is it a common
>practice, etc, so I'm looking for evidence, more than just "The Cisco TAC told
>me to do it".
>
>thanks
>
> Much appreciated, Eric
>_______________________________________________
>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/
>