[Xorp-users] PIM-SM in xorp cvs

[Atanu Ghosh]

Hi,

I am having difficulty understanding the problem my comments/questions
are inline below.

>>>>> "Otto" == Otto Solares <solca at guug.org> writes:

    Otto> On Tue, Aug 22, 2006 at 09:30:19AM -0700, Pavlin Radoslavov
    Otto> wrote:
    >> > FYI I fetch latest CVS and tried PIM-SM multicast routing and
    >> BGP > but BGP still don't respect the next-hop parameter so still
    >> using > quagga and trying xorp for PIM-SM:
    >> 
    >> Forgot to ask.  Can you provide more details about the issue with
    >> the XORP BGP next-hop parameter?

    Otto> As I report sometime ago, I can import nicely all 10,000
    Otto> routes into XORP via BGP, but I have 2 problems:

    Otto> * My setup is iBGP (my ASN and my peer ASN is the same), XORP
    Otto> say another router have this ASN.  If is iBGP it should accept
    Otto> this condition.

How exactly is this failure manifesting itself?

    Otto> * When peering with the route-reflector router (iBGP) XORP
    Otto> sends the messages with next-hop = some primary address in
    Otto> another NIC, it should send as PIM-SM works which sends it's
    Otto> messages using the "primary" address towards the RP, just in
    Otto> this case it should use the address toward the BGP router, In
    Otto> quagga this is achieved with:

    Otto> neighbor 10.10.26.14 update-source 10.10.26.7

    Otto> XORP is using 168.234.203.2 as source, quagga is using
    Otto> 10.10.26.7

When sending a message on an internal peering the NEXT_HOP should not be
rewritten (unless the route was internally generated). If XORP is
rewritting the NEXT_HOP this is incorrect. It seems that in your quagga
example that you are explicitly setting the NEXT_HOP. If you want to set
the NEXT_HOP values in XORP then you should be able to use policy to
rewrite the NEXT_HOP, there is an example on Page 100 of the user manual
that should provide hints.

    Otto> * Linux have notion of primary address and secondary addresses
    Otto> (ip a), XORP chooses the first numerically ordered IP address
    Otto> in an interface to be the primary address for sockets and
    Otto> stuff, it should pick the Linux primary address, this can be
    Otto> seen in xorpsh when issuing 'show * interface address".  This
    Otto> is not a serious problem as one can rearrange the IPs in your
    Otto> net so the first numerically address match the Linux primary
    Otto> address but anyway is a problem.

When the BGP rules dictate that a NEXT_HOP should be rewritten it will
always uses the configured value, never the first address on an
interface. When an address needs to be choosen we have tried very hard
to require the user to specify it.

    Otto> Hope I have been clear this time.

I think that we are getting closer.

  Atanu.