Routing flaps, was Re: Ping flooding

Mailman List Mirror (Read Only)
1

I'm sorry, but I'm afraid that you're missing the point here.

In most larger ISP backbones, the behavior of their IGP is indeed
visible to the public, since in most instances, most of the Internet
traffic relies on the stability of these interior (an esoteric term)
networks. Therefore, whether interior or exterior flap is really of
no relevance in this context.

In smaller ISPs, where not a great deal of public traffic is transiting
their (interior) backbone, granted, it is of lesser importance and
visibility.

- paul

2

I'm sorry, but I'm afraid that you're missing the point here.

In most larger ISP backbones, the behavior of their IGP is indeed
visible to the public, since in most instances, most of the Internet
traffic relies on the stability of these interior (an esoteric term)
networks. Therefore, whether interior or exterior flap is really of
no relevance in this context.

<This isn't intended to irritate, I just want to make sure I understand
this correctly>

Let me try a very simple example:

   To Internet To Internet <-- Peering points or upstreams
    'Point A' 'Point B'
        > >
  +-----------+ +-----------+
  > Router A |---| Router B |
  +-----------+ +-----------+
        > >
        > >
  +-----------+ +-----------+
  > Router C | | Router D |
  +-----------+ +-----------+
        > >
  To Downstream To Downstream
Non-BGP (Static) Non-BGP (Static)
    Customer Customer

Routers A & B are running BGP to the outside world, and iBGP between them.
Routers C & D are 'defaulted' into A and B.

If the link between A&B dies the Exterior Routes will (and should) flap.

If the link between either A or B and the Internet dies, the Exterior
Routes will (and should) flap.

If the line between routers A&C or B&D or between either C or D and their
respective static downstreams die, there should be NO external route
flap. However, if C&D are incapable of 'Null0' routing, it may be
beneficial to run dynamic routing between A&C and between B&D so that
A&B discard packets instead of causing a routing loop. This "internal
routing flap" should not be visible to the outside world.

I think that's what I meant to say before. Sometimes I'm not too clear
about what I'm saying. If I've still missed the boat feel free to let me
know :slight_smile:

-forrestc@imach.com

3

Forgot to throw one more item in here...

I'm sorry, but I'm afraid that you're missing the point here.

In most larger ISP backbones, the behavior of their IGP is indeed
visible to the public, since in most instances, most of the Internet
traffic relies on the stability of these interior (an esoteric term)
networks. Therefore, whether interior or exterior flap is really of
no relevance in this context.

In smaller ISPs, where not a great deal of public traffic is transiting
their (interior) backbone, granted, it is of lesser importance and
visibility.

What I forgot to say is that I fully agree with what you are saying here-
I doubt that a large provider such as Sprint or MCI or Agis or any other
major internet backbone is going to be able to hide much of their
internal route flapping, as their internal routes ARE their external routes.

My original note was intended to reply to the specific case of using
dynamic routing to distribute routes which could just as easily be done
with static routes - I.E. where there is one and only one path to the
destination. In this case, it is sometimes desirable and necessary to
run some sort of interior routing protocol, but it is not desirable that
flaps caused by a break along the one and only one path be propogated to
the outside world, but instead the packets should be null0'd at the point
of path convergence.

-forrestc@imach.com

4

==>If the line between routers A&C or B&D or between either C or D and their
==>respective static downstreams die, there should be NO external route
==>flap. However, if C&D are incapable of 'Null0' routing, it may be
==>beneficial to run dynamic routing between A&C and between B&D so that
==>A&B discard packets instead of causing a routing loop. This "internal
==>routing flap" should not be visible to the outside world.

It can be, though. Or at least from what I've seen.

If you have a 'floating static' to null0, it won't take over until the
dynamic holddown timer expires (whatever it happens to be for the
particular IGP you're using).

For example, a situation like this:

Router A <----> Router B <----> TermServ (with ISDN)<--DIAL-UP-LINK-->Router C

Router A is the access-point router doing BGP. The customer on router C
has a set of addresses assigned "way-back-when" by the InterNIC, and
router A is advertising that set via BGP.

Routing is done between A & B using OSPF. The LAN between B & TermServ is
running RIP (because certain manufacturers' boxen can't do anything BUT
RIP).

The organization is designed in such a way that their static dial-up
address is portable among all POPs in their provider (so the B & TermServ
could be in any of their POPs).

Router A has a static route to null0 to hold the BGP route in place. When
router C is dialed up and present, the route gets propogated from TermServ
to B via RIP, which redists into the OSPF area. Router A picks up this
route via OSPF.

The dynamically-learned route for customer 'C' now takes precedence over
the static route to null0 on A.

If for some reason, C drops the dial-in links and the TermServ goes
through a few RIP update cycles, the route will be marked as
'inaccessible', but will still be in the routing table until hold-down
expires. When this route is marked as 'inaccessible', the static route
does not take precedence and the route is withdrawn from BGP--and
consequently re-introduced when the hold-down expires and the null0 route
takes precedence.

Now, if there's something in the config of router A that can be used that
will prevent this flapping, or if it was just a freak coincidence that
routers saw a flap after dropping the connection, then please correct me.

/cah