Re: BGP advice: Customer converting from static ISP connection to BGP

["Christopher A. Woodfield" <rekoil () semihuman com>]

There really won't be a "blackhole" period if you do it right - if it's 
the same network that's currently being statically routed, the easy thing 
to do is to keep that static route until the BGP connection is up. There 
shouldn't be any routing interruption during the time that both are active 
- there will just be two announcements with different origin ASes. 

The hard part is ensuring that the BGP prefix has propagated to the 
internet as a whole before you pull the static route and as a result the 
ISP-sourced prefix - the best thing to do is to check looking glasses 
(make sure you check several - some networks will not forward the "new" 
route if the old one has a better metric) to ensure propagation of the new 
route before you pull the static route.

BGP routes typically take a few minutes to propagate globally, but this is 
a case where as long as the previous origin AS - the ISP - properly 
forwards traffic to the destination after the static route is pulled 
(which it should, if the new BGP peering is active), there shouldn't be 
any problems; at worst there will be some odd traceroutes for a few 
minutes.

-Chris

On Thu, Oct 25, 2001 at 04:38:48PM -0400, CONWAY () pjm com wrote:
> I'm looking for some independent confirmation from someone experienced in
> converting from a static routed ISP connection with the customer's netblock
> announced under the ISP's ASN to a BGP config where the customer has their own
> ASN and netblock.  We're the customer, and we're getting conflicting info on
> what to expect during the conversion.  
>
> Specifically, we're hearing conflicting things about how long it will take for
> the "world" to recognize our routes, whether or not there may be areas of the
> net "blackholed" from us for a while until things stabilize, etc.  
>
> I realize this is probably off-topic for this list, so please direct any replies
> directly to me.  If there is interest, I'm willing to summarize to the list.
>
> Chuck Conway
> conway () pjm com

-- 
---------------------------
Christopher A. Woodfield                rekoil () semihuman com

PGP Public Key: http://pgp.mit.edu:11371/pks/lookup?op=get&search=0xB887618B