[c-nsp] BGP MSS=576 bytes

[Church, Charles]

Is  "ip tcp path-mtu-discovery"  in the global config? 


Chuck 

-----Original Message-----
From: cisco-nsp-bounces at puck.nether.net
[mailto:cisco-nsp-bounces at puck.nether.net] On Behalf Of Antonio M.
Soares
Sent: Wednesday, February 11, 2009 10:36 AM
To: cisco-nsp at puck.nether.net
Subject: [c-nsp] BGP MSS=576 bytes


Hello group,

I have a 6500 running 122-18.SXF7 with lots of BGP peers and all of the
BGP sessions have negotiated a MSS of 536 bytes. Here's an
example:

++++++++++++++++++++++++++
6500>sh ip bgp neighbors x.x.x.x

...

Datagrams (max data segment is 536 bytes):

Rcvd: 439340 (out of order: 252), with data: 406672, total data bytes:
94316052

Sent: 296303 (retransmit: 727), with data: 35046, total data bytes:
994215

6500>
++++++++++++++++++++++++++

The documentation says that PMTUD is enabled by default so this should
not be happening:

++++++++++++++++++++++++++
BGP Neighbor Session TCP PMTUD

TCP path MTU discovery is enabled by default for all BGP neighbor
sessions, but there are situations when you may want to disable
TCP path MTU discovery for one or all BGP neighbor sessions. While PMTUD
works well for larger transmission links (for example,
Packet over Sonet links), a badly configured TCP implementation or a
firewall may slow or stop the TCP connections from forwarding
any packets. In this type of situation, you may need to disable TCP path
MTU discovery. In Cisco IOS Release 12.2(33)SRA,
12.2(31)SB, 12.2(33)SXH, 12.4(20)T, Cisco IOS XE Release 2.1, and later
releases, configuration options were introduced to permit
TCP path MTU discovery to be disabled, or subsequently reenabled, either
for a single BGP neighbor session or for all BGP sessions.
To disable the TCP path MTU discovery globally for all BGP neighbors,
use the no bgp transport path-mtu-discovery command under
router configuration mode. To disable the TCP path MTU discovery for a
single neighbor, use the no neighbor transport
path-mtu-discovery command under router or address family configuration
modes. 
++++++++++++++++++++++++++

I have for example a direct eBGP peering over TenGiga interfaces where i
see the same problem:

++++++++++++++++++++++++++
6500>sh int tenGigabitEthernet x/x | inc MTU
  MTU 1500 bytes, BW 10000000 Kbit, DLY 10 usec, 
6500>
6500>
6500>sh ip int tenGigabitEthernet x/x | inc MTU
  MTU is 1500 bytes
6500>
++++++++++++++++++++++++++



Any explanation to this strange behavior ?


Thanks.

Regards,

Antonio Soares, CCIE #18473 (R&S)
amsoares at netcabo.pt

_______________________________________________
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/