RE: TE metric and graceful restart

[Kireeti Kompella <kireeti@juniper.net>]

On Fri, 26 Apr 2002, Naidu, Venkata wrote:

>   1. First of all, your suggestion might not fit well
>      with draft draft-lefaucheur-te-metric-igp-01.txt.
>      That is, just changing TE metric might help you to
>      stop _some_ unwanted LSPs being setup in restart
>      period but not all. (example, LSPs that use IGP metric
>      in OSPF regular LSAs)

Turn that around: Yakov's proposal helps in all the cases that can
be helped.  Do you have a proposal to stop/discourage the setup of
LSPs that use the IGP metric?

>   2. If we change any of TE LSA contents and flood the
>      *Restart TE LSAs* to neighbor nodes, then there are
>      2 cases:
>
>      (a) The neighbor might have accepted as a helper
>          node
>      (b) The neighbor might have rejected (not at
>          all received any OSPF-grace LSA etc - for so
>          many reasons I can think of)
>
>    In the case (b), OSPF neighbor will simple flood the restart
>    TE LSA into the domain. That is something what you are not
>    intended in the below para:
>
>    ***
>    Neighbors of the restarting node should continue advertise the actual
>    unreserved bandwidth on the TE links from the neighbors to that node.
>    ***

You misread.  The TE links *from the node to the neighbors* change, not
the other way around.  By the way, the so-called restart TE LSAs are
intended to be flooded in all cases, not just when the neighbor doesn't
help.

>   3. More over, I think this is a Re-optimization issue rather than
>      a restart issue. So, IMHO, it is better not to flood the TE LSA
>      (with BW 0 and Infinite Metric changes).

Explain, please.

>   4. Actually, the LSP preventing in the grace period should
>      be handled in signaling protocols and definitely not
>      using back doors of routing protocols.

Explain, please.

Kireeti.