[Xorp-users] Well - I'm half working....

[Pavlin Radoslavov]

> 
> Hi Pavlin, others,
>  
> > 
> > Just curious, what is the reason you are running two routing engines?
> > 
> 
> At the core of our network we've got a Cisco Catalyst 6513 chassis with dual
> supervisor engines which take care of L2, management, etc. One supervisor is
> active, while the other waits for a failure. 
> 
> Each supervisor engine has it's own router, and to enable automatic failover
> for our L3 networks, we've deployed HSRP (Hot Standby Router Protocol),
> which also allows us to run both routers in an active-active arrangement,
> with automatic failover if a fault occurs.
> 
> Cheers,
> Hefin
> 
> HSRP -
> http://www.cisco.com/en/US/tech/tk648/tk362/tk321/tech_protocol_home.html

Interesting.

I did some search and I found the following from the cisco's web
site:
"Why Doesn't PIM Sparse Mode Work with a Static Route to an HSRP Address?"
http://www.cisco.com/warp/public/619/hsrpmcast.html

The text there describes exactly the same problem you have:

"In the example, the RPF neighbor is 10.1.3.3, which is the HSRP
standby address used by the default static route. However, this
address is not listed as a PIM neighbor. The reason the HSRP standby
address is not listed as a PIM neighbor is because the two routers
running HSRP (Routers 2 and 3) will not source the PIM neighbor
messages from the HSRP standby address."


The Web page suggests two solutions (and recommends the first one):

1. Run EIGRP on the downstream PIM router (i.e., the XORP router in
   your topology). Thus the downstream router will learn
   automatically the appropriate upstream RPF address.
   However, in your case you say that you don't want to run an
   unicast routing protocol (e.g., RIP) between the XORP router and
   the Cisco.

2. Add static multicast routes to the downstream router to make it
   RPF to the real routers' IP addresses.

   This was the solution I suggested, but of course it has the drawback
   that the RPF will stop working if the engine with the RPF address
   fails.
   However, at this moment this is the only solution you have,
   unless you decide to run RIP between the cisco router and XORP.


While on the subject. Today (independently from the HSRP issue) we
had a discussion about importing unicast RIP/BGP/OSPF routes into
the MRIB at the protocol-level granularity if you are running one
of those protocols (currently, the only solution is to use fib2mrib
which imports all unicast forwarding entries).
At the end of the discussion we concluded that we need a mechanism
at each protocol's output. That mechanism can be configured such
that the route add/delete commands that are typically sent only to
the URIB can be sent to the MRIB as well.
However, once we implement that mechanism, then it will be trivial
to configure it such that the add/delete commands from, say, RIP are
sent only to MRIB. In your case, this is exactly what you want:
you can run RIP between the Cisco and XORP, but those routes won't
affect the unicast routing in your XORP multicast router.
Hopefully, we will have this functionality for the 1.1 release.
In the mean time I'd recommend that you use the static mrib-route4
setup to one of the reals IP addresses.

Regards,
Pavlin