Re: GMPLS/RSVP-TE question

[Suresh Katukam <skatukam@cisco.com>]

Cheenu,

Let me try to answer your questions to best of my knowledge..

Cheenu Srinivasan wrote:
> 
> Reading the GMPLS drafts I have a question about setting up a
> generalized LSP with RSVP-TE. Consider the following network where
> the control plane is distinct from a data plane which is IP unaware.
> 
>      D --- E
>    /   \ /   \
>   A =1= B =2= C
> 
>  --- represent control plane links
>  =n= represent data plane links which cannot originate/terminate IP;
>      n is the local ID for this link
> 
> We want to set up an LSP A-1-B-2-C and signal it via RSVP-TE. The ERO
> should be (A-1, B-2, C). However, a Path message can only be sent
> out link A-D and has to be routed towards B; it cannot be sent
> down link 1 since 1 is not IP-capable. Is this ok? 

Yes, This is completely OK. This is the reason why OIF has introduced
OF/OB  (Out-of-Fiber/Out-of-Band) where the data links may not be
able to terminate packets/understand IP.

> What kind of
> Path state is created on D? Or, do we need to tunnel the Path
> message to B directly? Will D change the PHOP object to point
> to itself (we do not want D to change the contents of the ERO)?

D and E should not be involved in creating LSP at all. These
should pass messages through. For this, A has to send a directed
message to B and then B has to send a directed message to C. 
There should not be any router alert option bit set - this avoids
D and E looking into RSVP packets.

> In
> this case when B receives this message it will find that the
> Path came to it from an interface that is not part of the ERO. Is
> this an error?

No. This is not an error. As long as you receive a PATH message on
an interface that will take you to the previous hop, I do not
think that it should cause ERO processing error. What you really
need to validate is that in ERO, do you have link that corresponds
to A and its interface ID as per draft on handling unnumbered interfaces
in RSVP.

Thanks,
Suresh