The MPLS WG Archive

[Eric Gray <eric.gray@sandburst.com>]

Katta,

    It is something of an implementation choice, however, opinions on a
'correct' model can be both variant and enthusiastic.  Therefore, it will
make sense to support either approach, optionally, in implementations.

    If you pre-suppose that B has established the alternate path (B-X-C)
in advance, then it must be clear that B assumes a correct behavior that
involves switching to the alternative path.  If you assume otherwise, then
there must be some time-bound criteria on which an operator would like
to see some modification in switching behavior that minimizes lost traffic
and wasted networking resources associated with forwarding packets to
B so that they can be discarded as a result of no extant forwarding path
downstream from B.

    I believe these are the rough parameters defining the domain in which
the two approaches are an implementation choice.

--
Eric Gray

You wrote:

> Hi,
> I have a doubt in Route pinning of CR-LDP.
>
> Suppose an LSP is pinned to a route. The draft says if there is a new better
> route for the established LSP, it should not be considered for pinned LSP.
>
> But if the route becomes completely un-available (say next-hop) failed and
> an alternate path which meets LSP requirements exist, what action needs to
> be taken. Does the CR-LSP should be teared down as pinning condition cant be
> met or the alternate path which meets the requirements should be considered
> to set-up LSP from failed point considering the view point that pinning was
> for avoiding disruption of services.
>
> problem explained with example:
>
> A,B,C,D is Explicit route specified, with C as loose hop.
> Suppose LSP is set-up with nodes A,B,Z,C,D and pinned.
> Now node Z failed, but there is alternate path B,X,C which meets LSP
> requirements.
> Now what action B should take when it detects it cant reach Z. Should it try
> to change LSP path as A,B,X,C,D or tear down the LSP.
>
> If it is implemetation choice, please give your analysis of pros and cons of
> each approach.
>
> Thanks in advance.
>
> regards,
> Katta