[c-nsp] Non-default BGP hold / keepalive timers

Bruce Pinsky bep at whack.org
Mon Oct 3 18:22:30 EDT 2005


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

David J. Hughes wrote:
> 
> On 04/10/2005, at 6:34 AM, Bruce Pinsky wrote:
> 
>> Start talking to your providers and peers about when they are going to
>> offer BFD support.
> 
> 
> Yup, in a perfect world.  However, I need a solution now, not next
> year.  And I'm sure that our government owned Telco will move as swiftly
> towards supporting BFD as they have towards other sane options (like
> supporting standard communities etc).  Doesn't happen around here.
> 


Yes, moving the gov't is like moving an elephant...sigh...

> 
>> This will be the preferred method since using layer 3
>> routing protocols as a fault detection tool has significant drawbacks.
> 
> 
> Expecting a dynamic routing protocol to dynamically route around a
> failed link isn't asking for too much in my opinion.  Using the features
> defined by the protocol's spec to determine when a path isn't stable
> would also appear reasonable.
> 

However a non-operational link presenting itself as an operational link is
not reasonable.   A dynamic routing protocol is designed to react to
changes in link availability and reachability as reported at layer 1 and
layer 2.  The timers of the routing protocol are designed to be failsafe
mechanisms for "non-well-behaved" failures not as a substitute for proper
reporting of link faults by the underlying media.

> 
>> I don't know that many providers are willing to tune the BGP timers on a
>> per customer basis at all.  YMMV.
> 
> 
> They don't need to.  As per the spec the peering routers will agree on
> the lower of the 2 hold times presented during session negotiation. 
> Keepalive times will be internally set to one third of the negotiated
> hold timer.  I've tried setting a reduce hold timer at my end and it's
> negotiated as expected with the provider in question.  The other party
> doesn't need to do anything.
> 
> 

Providing your BGP peering neighbor's implementation doesn't have knobs to
disallow a connection below a certain holdtime value.

- --
=========
bep

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.0 (MingW32)

iD8DBQFDQa8mE1XcgMgrtyYRAnIpAKDaGZNbZVNi+Bcf0UOW5TK+6JeIGQCgwHw/
QoH5idakw4LuFCE1KcySd9k=
=nrSM
-----END PGP SIGNATURE-----


More information about the cisco-nsp mailing list