The MPLS-OPS Archive

Cell Relay Retreat>MPLS-OPS Archive>month:2004-Sep> msg00149



[Date Prev][Date Next][Thread Prev][Thread Next]  
  [Date Index][Thread Index][Author Index][Subject Index]

Re: IP Routing problem after enabling MPLS?

  • From: Garry Glendown <garry@regio.net>
  • Date: Wed, 22 Sep 2004 09:00:36 +0200
  • Resent-Date: Wed, 22 Sep 2004 03:31:54 -0400
  • User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.6) Gecko/20040113

Abhishek Mittal wrote:

 > Hi Garry,
 >
 > What do you mean by Dialup (DSL) customers.
 >
 > As per your network diagram , your DSL customers are getting 
connected on Cisco 2900 switch & this L2 switch is connected to your PE3.


Dialup comes through PE3 which gets the DSL Dialups via an L2TP tunnel 
from the DSL backbone ...

With the same config that I listed in the previous mail, connections 
into our net work fine when MPLS is disabled on PE3's FE, and stop 
working when MPLS is enabled ... which makes me wonder, as the MTU is 
WAY below the MTU used on the backbone - so why doesn't it work once 
MPLS is active?

just a couple more results of some tests I did ... don't know whether 
any of those will help ...

pings with packet sizes of up to 1456 get through from the customer to 
any destination server when I have mpls disabled ... (ping -s 1428)

                  +------SWITCH-----+
                  |        |        |
       Uplink----PE1      PE2      PE3-----Customer Dialup (DSL L2TP)
                         /  |
                  PEERING   | /-----P3---P4---W2
                            |/
                    PE5-----P1-----Switch-------W1
                            |         |
                            | E3      | FE-Bridge
                            |         |
                  PE6--P2--PE4-----Switch
                   |        |
                 Switch     CE
                   |
                   W3

Now to something weird ...


I just did a tcpdump on an FTP session ... with MPLS disabled, I see 
these packets (among others  ):

08:28:52.597101 customer.23606 > server.ftp: S 1454177743:1454177743(0)
     win 5840 <mss 1416,sackOK,timestamp 26546056 0,nop,wscale 1>
                       ^^^^
This is what is expected, after all I set the adjust-mss value to 1416 ...

But with MPLS enabled (no other changes at all), I see this:

08:26:08.241047 customer.23603 > server.ftp: S 1279761092:1279761092(0)
     win 5840 <mss 1432,sackOK,timestamp 26381662 0,nop,wscale 1>
                       ^^^^

Why is mss set/forced to this just because MPLS routing is active? Also, 
changing the mss-adjust to anything else still results in the mss of 
1432 arriving at the server ... !? Is this documented behaviour?

In order to compensate for this mss, I have now altered the MTU on the 
dialup connections from 1456 to 1472 - now the connections I've tried 
and couldn't get to work before work fine !? I will need to wait if the 
customer who complained yesterday is still working, but I suppose he is ...

Any insights?

Tnx, -gg

-------
The MPLS-OPS Mailing List
Subscribe/Unsubscribe:  http://www.mplsrc.com/mplsops.shtml
Archive: http://www.mplsrc.com/mpls-ops_archive.shtml