|
Neohapsis is currently accepting applications for employment. For more information, please visit our website www.neohapsis.com or email hr@neohapsis.com |
From: Sina Mirtorabi (sina
CISCO.COM)Date: Sat Jun 09 2001 - 14:10:30 CDT
Pat,
"Pat Murphy - (650)329-4044" wrote:
> Sina,
>
> Your analysis is correct. Indeed try imagining that
>
> the ABR is learning some external route over an interface which
> is not internal to the NSSA but is internal to the OSPF domain.
>
> You still have to select a forwarding address which is internal to the
> NSSA (See RFC 1587 Section 3.5 Step 3). Clearly routing may be
> suboptimal in this case. Sometimes you have to give a little to get a
> lot. In the latest Type 5 to Type 7 Translation draft, it is desirable
> to allow, under certain restrictions, inter-area forwarding addresses
> for Type 7 LSAs originated by an NSSA ABR. Unfortunately this feature is
> not backward compatible.
>
> Pat
I don't really agree with your below statement ;-)
=========
Indeed try imagining that
the ABR is learning some external route over an interface which
is not internal to the NSSA but is internal to the OSPF domain.
You still have to select a forwarding address which is internal to the
NSSA (See RFC 1587 Section 3.5 Step 3)
=========
here is section 3.5 step 3 I don't see where it's stated
--------------
3. Now, look up the routing table entry for the destination
N. If no entries exist for N, install the AS external path
to N, with the next hop equal to the list of next hops to
the forwarding address/ASBR, and the advertising router equal
to ASBR. If the external metric type is 1, then the
path-type is set to Type-1 external and the cost is equal
to X + Y. If the external metric type is 2, the path-type
is set to Type-2 external, the link-state component of the
route's cost is X, and the Type-2 cost is Y.
------------
but its stated ( at least in your latest NSSA draft ) if the interface
over which we learn the route is not internal to NSSA a forwarding address
should be set from one of the interface belonging to NSSA ( in the RFC 1587
" belonging to NSSA'' is not mentioned )
my whole point is "for NSSA ABR the forwarding address of type 7 should be
zero". here is the reason
a) an NSSA ABR generating external information ( type 5/7) must clear the P
bit for type 7 LSA generated, so this LSA will not be translated to type 5
and we don't violate the rule " all the type 7 LSA with forwarding bit set
must have a forwarding address "
Further by setting to forwarding address to one of the router's interface in
NSSA area, we can create Non-optimal routing as described before ( other
NSSA routers will sent packet to this address which is chosen randomly and
can not be the optimum path and by setting to zero they will find the best
intra area path to ASBR )
b) the whole idea of setting the forwarding address was
- to have an optimal routing in case of having more than one NSSA ABR and
type 7 to 5 translation ( P bit set )
here its definitely not the case as the P bit is cleared and the ABR is
originating itself the external route
- ease of transition from type 7 to type 5 as we will copy the forwarding
address
here there is no ease of transition as the ABR is generating itself both
type 5 and type 7 and besides the forwarding address should not be the same
for type 5 and 7
for example imagine an NSSA ABR learning over an multi-access interface
which is internal to OSPF but external to NSSA area, some external
information. we have to set the forwarding address for type 5 to the next
hop over which we learned the external in order to have optimal routing for
type 5 BUT we can't have this address in type 7 as this is not internal to
NSSA.
if we set the forwarding address for type 5 and type 7 to one of the NSSA
area interface not only we would have Non-optimal routing for NSSA area as
described in point a) But now even type 5 will have Non-optimal routing as
we are not using the next hop over which we learned the route
so clearly type 5 and type 7 generated by a NSSA ABR should be considered
separately
So to summarize all the above, if a NSSA ABR generates type 5 and type 7,
those should be considered independently, type 5 forwarding address will
follow its rule described in the RFC and the type 7 LSA should have its
forwarding address set to ZERO in order to avoid Non-optimal routing
after all the whole idea of setting the forwarding address was to avoid
Non-optimal routing but here for NSSA ABR we introduce Non-optimal routing
so the above statement make sense.
lastly if we do this we can suppress the below phrase in your latest draft.
since by doing this we can introduce Non-optimal routing even for type 5
" if a router ( ABR ) originates a type 5 LSA with a forwarding address
part of NSSA, it should also originate a type 7 into NSSA area"
thanks
Sina
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]