[c-nsp] MPLS LDP neighbouship Flapping

[Aamer Akhter (aakhter)]

Atiqur,

The following config is allowing LDP to expand to a higher MSS:

ip tcp path-mtu-discovery

But there are some cases where the DF=1 setting will not always generate an ICMP unreachable, which would have allowed LDP/TCP to tune the MSS down to something acceptable.

As a VPLS situation will not allow you to do fragmentation at the edge (and even in the L3VPN case PE-fragmentation is not really something you want to do), you really do want to increase your core MTU to a higher value than your attachment circuits + any tunneling overhead.



-- 
Aamer Akhter / aa at cisco.com
NSITE - cisco Systems 

> -----Original Message-----
> From: cisco-nsp-bounces at puck.nether.net 
> [mailto:cisco-nsp-bounces at puck.nether.net] On Behalf Of 
> Atiqur Rahman Mohammed
> Sent: Wednesday, November 16, 2005 9:57 AM
> To: cisco-nsp; mpls-ops at mplsrc.com
> Subject: [c-nsp] MPLS LDP neighbouship Flapping
> 
> Please find the resolution to issue. The problem is resolved  by TAC
> case .  Provide your comments.--
> Regards,
> Atiqur Rahman
> Infocomm Technology Cente
> Reliance Infocomm
> Mobile: 09324621784
> 
> 
> desakov at cisco.com
> 11/16/2005 07:58 PM	
> To
> atiqurrahman.mohammed at relianceinfo.com
> cc
> 
> Subject
> 602346373 Packet loss between two connected c7600 -  
> CT20051029_0000001364
> 	
> 
> Hi Atiqur,
> 
> The issue with LDP neighborship and EoMPLS VCs is caused
> by the fact, that 1500 bytes ip packets can not go from Loopback
> 192.168.71.165 to Loopback  192.168.71.161. As you said, only
> 1470 bytes packets can travel accross the "Core" network.
> 
> I could see that TCP session (LDP uses TCP) between 192.168.71.165
> and 192.168.71.161 gets established, but when it comes to label
> advertisements, TCP keeps retransmitting the packets and eventually
> times out.
> 
> So to make LDP working, we shall provide it with sufficient MTU.
> Now I configured 'ip mtu 1462' on the interface of 
> "192.168.71.165" and
> this caused TCP to reduce it's packet size and now LDP session is
> stable and EoMPLS VCs are up.
> 
> However, I do not think this is a solution, because user 
> traffic in this
> EoMPLS VCs will be 1500 bytes of IP packet + ~18 bytes ethernet
> header + MPLS label. This will not work if the maxim packet you can
> send accross "Core" is 1470. You shall check with administrators
> of "Core" to fix the MTU issue. Normally, you shall be fine if you
> can send accross 1530 bytes MPLS packets.
> 
> Please let me know if there are any questions or if there is anything
> else what I can do for you. Thanks!
> Dmitriy
> 
> Cisco Technical Support Website: http://www.cisco.com/techsupport
> Cisco TAC Service Request Tool: 
> http://www.cisco.com/techsupport/servicerequest
> Cisco Technical Support Newsletter: 
> http://www.cisco.com/techsupport/newsletter
> 
> _______________________________________________
> 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/
>