[c-nsp] BGP MSS=576 bytes

[Mark Tinka]

On Thursday 12 February 2009 10:33:39 am Seth Mattinen 
wrote:

> Just for fun I looked at my BGP mesh and found the
> following:
>
> * Directly connected neighbors MTU=516
> * Transit to upstreams MTU=1440 or 1460
>
> This is under 12.4(16) and I've never touched 'ip tcp
> path-mtu-discovery' or per-neighbor MTU settings.

From our busier route reflectors (7201's), we are seeing:

* For v4, some of our IOS routers are agreeing on 1,440
  bytes, while others on 8,916 bytes. Need to look into
  the difference here - I recall we recently replaced a
  defective route reflector, which is the one registering
  the higher MTU.

* For v6, all our IOS routers are agreeing on 8,896 bytes.

Interface MTU is 9,000 bytes on all router interfaces. IOS 
code is 12.2(33)SRC3.

We have nearly similar values for our JunOS routers peering 
with the route reflectors, but not sure that's relevant 
here.

Suffice it to say that 'ip tcp path-mtu-discovery' isn't 
configured, although it is enabled by default for BGP:

#sh ip bgp neighbors
<snip>
Transport(tcp) path-mtu-discovery is enabled

#sh bgp ipv6 unicast neighbors
<snip>
Transport(tcp) path-mtu-discovery is enabled

For routers where we've configured 'no bgp transport path-
mtu-discovery' under the BGP process (for 
troubleshooting/testing), we've registered an agreed value 
of 516 bytes.

eBGP sessions to downstreams is at 1,440 bytes.

Cheers,

Mark.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 835 bytes
Desc: This is a digitally signed message part.
URL: <https://puck.nether.net/pipermail/cisco-nsp/attachments/20090212/0067bc11/attachment.bin>