OSPF Fast Hello

Michael Vallaly bird at nolatency.com
Fri Jun 29 00:35:22 CEST 2012


Greetings fellow bird users,

Figured I would resurrect this thread as I am running into similar
issues with Bird 1.3.7 and OSPF across a few gigabit fiber PtP links
in Linux (3.3.6).

The goal is to provide the fastest reliable convergence of OSPF when
experiencing an upstream (L3) link failure (OSPF over GRE tunnels).   

What are the realistic minimum/fastest hello settings bird supports? 

I have noticed that setting the hello <= 2 (even on local gigE
segments) causes the OSPF process to constantly go down and re-init
roughly every 15 seconds causing lots of CPU load and constant route
flaps.

Cisco seems to provide a mechanism/feature for Fast Hellos
(http://www.cisco.com/en/US/docs/ios/12_0s/feature/guide/fasthelo.html),
is this something Bird supports or plans to support?

Currently my interfaces look something like this:

<snip>
interface "ptp1" {
 type ptp;
 hello 5; retransmit 10; dead count 2; wait 5;
};
</snip>

Am I doing something obviously wrong, or expecting too much out of OSPF?

Thanks!

-Mike 

On Tue, 22 May 2012 11:58:07 +0000 (GMT)
Javier Lorenzo Prieto <javier.lorenzo at me.com> wrote:

> Hello,
> 
> I tested in the lab with 2 links ethernet and the result is the same: if hello tiem is less than 3 do not set the adjacency.
> 
> With Quagga software or Cisco routers, is possible. Is it possible to achieve rapid recovery using Bird?
> 
> Thanks in advance for your help!
> 
> El 20 may 2012 a las 14:00, Henrique de Moraes Holschuh <hmh at hmh.eng.br> escribió:
> 
> On Sat, 19 May 2012, Javier Lorenzo Prieto wrote:
> > We are using BIRD to configure load balancing and redundancy between two
> > sites. These sites are interconnected via two short-range RF links (each
> > with 15 Mpbs effective throughput and delay of 5ms). Because we intend to
> > send VoIP over these links and would like to avoid dropping calls in the
> > event of interruptions in one of the links. Therefore, dynamic routes need
> > to be updated very quickly.
> 
> How does multicast/broadcast frames work on that RF link?
> 
> If it is a Wifi-like link (i.e. it uses a 802.11 MAC), you most likely have
> to avoid anything non-unicast like the plague. Please try OSPF in NBMA mode.
> 
> -- 
> "One disk to rule them all, One disk to find them. One disk to bring
> them all and in the darkness grind them. In the Land of Redmond
> where the shadows lie." -- The Silicon Valley Tarot
> Henrique Holschuh


-- 
Michael Vallaly <mvallaly at nolatency.com>



More information about the Bird-users mailing list