The MPLS-OPS Archive

Cell Relay Retreat>MPLS-OPS Archive>month:2001-Oct> msg00202



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

RE: Fragmentation in MPLS network BGP problem

  • From: Marcelo Blanes <mblanes@larc.usp.br>
  • Date: Wed, 24 Oct 2001 09:55:15 -0200 (EDT)
  • cc: Robert Raszuk <raszuk@cisco.com>, mpls-ops@mplsrc.com
  • Resent-Date: Wed, 24 Oct 2001 10:39:47 -0400
  • To: Alexander Marhold <alexander@marhold.at>
  • X-Sender: mblanes@melao


Does BGP try to perform MTU path-discovery? Do I need to worry regarding
fragmentation problems?

Thanks,
Marcelo.

On Wed, 24 Oct 2001, Alexander Marhold wrote:

> I have seen a similar thing therfore my questions
> 
> are you using 26xx and7or 36xx routers in you network ?
> 
> as those are not fragmenting a packet when the unlabeled packet is below the
> MTU size but the labelled packet is above.
> It seems that the first fragmentation works giving a 1496 byte frame, but
> the answer does not reach the source.
> 
> However as far as I know beside PING and TRACEROUTE all applications are
> setting the DF-bit and are using path-MTU discovery mechanism.
> 
> so maybe try the same thing with DF set and look what messages are comming
> back and from which node
> 
> with best regards
> 
> Alexander
> 
> __/__/__/__/__/__/__/__/__/__/__/__/__/__/__/__/__/__/__/__/__/
>      __/
>     __/   Dipl.Ing. Alexander Marhold MBA
>    __/   CCIE #3324, CCNP, CCDP, CCSI #20642
>   __/   Core & IP Services <Senior Consultant>
>  __/   Mobile: ++43-(0)664-16 28 234
> __/    PRO IN http://www.proin.com
> 
> 
> 
> -----Original Message-----
> From: Marcelo Blanes [mailto:mblanes@larc.usp.br]
> Sent: Tuesday, October 23, 2001 8:55 PM
> To: Robert Raszuk
> Cc: mpls-ops@mplsrc.com
> Subject: Re: Fragmentation in MPLS network BGP problem
> 
> 
> 
> Robert,
> 
> See my pings bellow:
> 
> For ICMP 1500 ok for 2000 it fails.
> 
> The 1500 ping...
> 
> PE01#ping
> Protocol [ip]:
> Target IP address: pe02 (loopback address)
> Repeat count [5]:
> Datagram size [100]: 1500
> Timeout in seconds [2]:
> Extended commands [n]: y
> Source address or interface: pe01 (loopback address)
> Type of service [0]:
> Set DF bit in IP header? [no]:
> Validate reply data? [no]:
> Data pattern [0xABCD]:
> Loose, Strict, Record, Timestamp, Verbose[none]:
> Sweep range of sizes [n]:
> Type escape sequence to abort.
> Sending 5, 1500-byte ICMP Echos to 10.190.107.251, timeout is 2 seconds:
> !!!!!
> Success rate is 100 percent (5/5), round-trip min/avg/max = 1/2/4 ms
> PE01#
> 04:45:01: IP: s=10.190.107.252 (local), d=10.190.107.251, len 1500, cef
> process switched
> 04:45:01:     ICMP type=8, code=0
> 04:45:01: IP: s=10.190.107.252 (local), d=10.190.107.251, len 1500, cef
> process switched
> 04:45:01:     ICMP type=8, code=0
> 04:45:01: IP: s=10.190.107.251 (FastEthernet12/0/0), d=10.190.107.252,
> len 1500, rcvd 4
> 04:45:01:     ICMP type=0, code=0
> 04:45:01: IP: s=10.190.107.252 (local), d=10.190.107.251, len 1500, cef
> process switched
> 04:45:01:     ICMP type=8, code=0
> 04:45:01: IP: s=10.190.107.252 (local), d=10.190.107.251, len 1500, cef
> process switched
> 04:45:01:     ICMP type=8, code=0
> 04:45:01: IP: s=10.190.107.251 (FastEthernet12/0/0), d=10.190.107.252,
> len 1500, rcvd 4
> 04:45:01:     ICMP type=0, code=0
> 04:45:01: IP: s=10.190.107.252 (local), d=10.190.107.251, len 1500, cef
> process switched
> 04:45:01:     ICMP type=8, code=0
> 04:45:01: IP: s=10.190.107.252 (local), d=10.190.107.251, len 1500, cef
> process switched
> 04:45:01:     ICMP type=8, code=0
> 04:45:01: IP: s=10.190.107.251 (FastEthernet12/0/0), d=10.190.107.252,
> len 1500, rcvd 4
> 04:45:01:     ICMP type=0, code=0
> 04:45:01: IP: s=10.190.107.252 (local), d=10.190.107.251, len 1500, cef
> process switched
> 04:45:01:     ICMP type=8, code=0
> 04:45:01: IP: s=10.190.107.252 (local), d=10.190.107.251
> PE01#, len 1500, cef process switched
> 04:45:01:     ICMP type=8, code=0
> 04:45:01: IP: s=10.190.107.251 (FastEthernet12/0/0), d=10.190.107.252,
> len 1500, rcvd 4
> 04:45:01:     ICMP type=0, code=0
> 04:45:01: IP: s=10.190.107.252 (local), d=10.190.107.251, len 1500, cef
> process switched
> 04:45:01:     ICMP type=8, code=0
> 04:45:01: IP: s=10.190.107.252 (local), d=10.190.107.251, len 1500, cef
> process switched
> 04:45:01:     ICMP type=8, code=0
> 04:45:01: IP: s=10.190.107.251 (FastEthernet12/0/0), d=10.190.107.252,
> len 1500, rcvd 4
> 04:45:01:     ICMP type=0, code=0
> PE01#
> 
> 
> The 2000 ping...
> 
> PE01#ping
> Protocol [ip]:
> Target IP address: pe02
> Repeat count [5]:
> Datagram size [100]: 2000
> Timeout in seconds [2]:
> Extended commands [n]: y
> Source address or interface: pe01
> Type of service [0]:
> Set DF bit in IP header? [no]:
> Validate reply data? [no]:
> Data pattern [0xABCD]:
> Loose, Strict, Record, Timestamp, Verbose[none]:
> Sweep range of sizes [n]:
> Type escape sequence to abort.
> Sending 5, 2000-byte ICMP Echos to 10.190.107.251, timeout is 2 seconds:
> .....
> Success rate is 0 percent (0/5)
> PE01#
> 05:00:51: IP: s=10.190.107.252 (local), d=10.190.107.251, len 2000, cef
> process switched
> 05:00:51:     ICMP type=8, code=0
> 05:00:51: IP: s=10.190.107.252 (local), d=10.190.107.251, len 2000, cef
> process switched
> 05:00:51:     ICMP type=8, code=0
> 05:00:51: IP: s=10.190.107.251 (FastEthernet12/0/0), d=10.190.107.252, len
> 504, rcvd 4
> 05:00:51:     IP Fragment, Ident = 46, fragment offset = 1496
> 05:00:51: IP: recv fragment from 10.190.107.251 offset 1496 bytes
> 05:00:53: IP: s=10.190.107.252 (local), d=10.190.107.251, len 2000, cef
> process switched
> 05:00:53:     ICMP type=8, code=0
> 05:00:53: IP: s=10.190.107.252 (local), d=10.190.107.251, len 2000, cef
> process switched
> 05:00:53:     ICMP type=8, code=0
> 05:00:53: IP: s=10.190.107.251 (FastEthernet12/0/0), d=10.190.107.252, len
> 504, rcvd 4
> 05:00:53:     IP Fragment, Ident = 47, fragment offset = 1496
> 05:00:53: IP: recv fragment from 10.190.107.251 offset 1496 bytes
> 05:00:55: IP: s=10.190.107.252 (local), d=10.190.107.251, len 2000, cef
> process switched
> 05:00:55:     ICMP type=8, code=0
> 05:00:55: IP: s=10.190.107.252 (local), d=10.190.107.251, len 2000, cef
> process switched
> 05:00:55:     ICMP type=8, code=0
> 05:00:55: IP: s=10.190.107.251 (FastEthernet12/0/0), d=10.190.107.252, len
> 504, rcvd 4
> 05:00:55:     IP Fragment, Ident = 48, fragment offset = 1496
> 05:00:55: IP: recv fragment from 10.190.107.251 offset 1496 bytes
> 05:00:57: IP: s=10.190.107.252 (local), d=10.190.107.251, len 2000, cef
> process switched
> 05:00:57:     ICMP type=8, code=0
> 05:00:57: IP: s=10.190.107.252 (local), d=10.190.107.251, len 2000, cef
> process switched
> 05:00:57:     ICMP type=8, code=0
> 05:00:57: IP: s=10.190.107.251 (FastEthernet12/0/0), d=10.190.107.252, len
> 504, rcvd 4
> 05:00:57:     IP Fragment, Ident = 49, fragment offset = 1496
> 05:00:57: IP: recv fragment from 10.190.107.251 offset 1496 bytes
> 05:00:59: IP: s=10.190.107.252 (local), d=10.190.107.251, len 2000, cef
> process switched
> 05:00:59:     ICMP type=8, code=0
> 05:00:59: IP: s=10.190.107.252 (local), d=10.190.107.251, len 2000, cef
> process switched
> 05:00:59:     ICMP type=8, code=0
> 05:00:59: IP: s=10.190.107.251 (FastEthernet12/0/0), d=10.190.107.252, len
> 504, rcvd 4
> 05:00:59:     IP Fragment, Ident = 50, fragment offset = 1496
> 05:00:59: IP: recv fragment from 10.190.107.251 offset 1496 bytes
> PE01#
> 
> Any light?
> 
> Thanks,
> Marcelo.
> 
> 
> On Sun, 21 Oct 2001, Robert Raszuk wrote:
> 
> > Marcelo,
> >
> > > I have PE - P - PE router. If I do a ping in PE to PE if datagram size
> for
> > > example 2000 (PE to P router has ether interface if tag mut 1526) with
> DF
> > > bit not set, I got I timeout answer. For 1500 it work fine.
> >
> > You would need to narrows down where your packets are being lost. Maybe
> > you have a bug of not bumping to process switching (some platforms like
> > GSR don't support that at all),
> >
> > > Ok Robert, but iBGP messages ( session VPNv4 ) is exchange between
> > > loopacks. Will iBGP looks to MTU in the loopback?!
> >
> > No it does not matter what it the IBGP update-src or dst. It matters via
> > what PHYSICAL interface the packets are going out of your router, IBGP
> > is not aware about any MTU issue. It just passed the generated updates
> > to TCP/IP stack and all fragmentatation is done there.
> >
> > R.
> >
> > -------
> > The MPLS-OPS Mailing List
> > Subscribe/Unsubscribe:  http://www.mplsrc.com/mplsops.shtml
> > Archive: http://www.mplsrc.com/mpls-ops_archive.shtml
> >
> 
> -------
> The MPLS-OPS Mailing List
> Subscribe/Unsubscribe:  http://www.mplsrc.com/mplsops.shtml
> Archive: http://www.mplsrc.com/mpls-ops_archive.shtml
> 
> -------
> The MPLS-OPS Mailing List
> Subscribe/Unsubscribe:  http://www.mplsrc.com/mplsops.shtml
> Archive: http://www.mplsrc.com/mpls-ops_archive.shtml
> 

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