Re: [mpls] working group last call on draft-ietf-mpls-ldp-igp-sync-bcast-00.txt

[Sriganesh Kini <[e-mail]>]

Inline @ [Sri] 


- Sri

-----Original Message-----
From: Mikael Abrahamsson [mailto:[e-mail]] 
Sent: Tuesday, March 02, 2010 11:36 PM
To: Sriganesh Kini
Cc: [e-mail]
Subject: RE: [mpls] working group last call on draft-ietf-mpls-ldp-igp-sync-bcast-00.txt

On Wed, 3 Mar 2010, Sriganesh Kini wrote:

> [Sri] For a p2p link that is typically what should happen (if RFC5443 
> is implemented correctly). For multi-access links see the draft 
> section 1 that explains what problems would occur.

This is mainly HDLC/PPP links over POS, or (ten)GE links configured with ISIS point-to-point.

When we set cost at one end, this only affects cost one way which is what I would like to change so that it would be bidirectional.

[Sri] This problem should not happen if RFC5443 is implemented correctly. Typically both ends of a p2p link will have IGP come up simultaneously but until LDP is established, both routers should be advertising max cost. 

However what might be happening is that the link may be declared up by physical layer at different times. This is typically true of p2p GiGE links if they are going through an intermediate switch & is also true of POS links going over certain types of L1 switches. Each end of the router may detect physical link up but due to the intermediate link between switches being down the IGP (& LDP) adjacency would not come up. After a certain timeout the link (stub) would be advertised with actual cost. Now later if the physical link comes up end-to-end between the two routers, IGP metric must be again changed back to max cost (on both sides) if RFC5443 is interpreted correctly. If this is done then there should be no issue, but there is potential to not get this right and could explain problems that you experienced. 


-- 
Mikael Abrahamsson    email: [e-mail]