[Xorp-users] route flap damping triggers on eBGPduplicate updates?

[Jonathan Park]

Hi Bruce,

I've just tested it, and it *does* penalize duplicates.
In fact as you said, it penalizes *all* updates with penalty = 1000.

Route flap damping RFC specifies that there should be a penalty for each
flap. 

>From my observation, different implementations interpret "flap" in different
ways.  Some implementations penalize all updates (like XORP), and some
implementation (Quagga) penalize an announcement following a withdrawal.
Cisco penalizes withdrawals and updates with attribute changes but not
duplicates.  Juniper penalizes all updates but not duplicates.

I was wondering whether this is a bug or it is by design.

Thanks,

Jonathan

> -----Original Message-----
> From: xorp-users-bounces at xorp.org [mailto:xorp-users-bounces at xorp.org] On
> Behalf Of Jonathan Park
> Sent: Saturday, May 09, 2009 1:05 PM
> To: Bruce Simpson
> Cc: xorp-users at xorp.org
> Subject: Re: [Xorp-users] route flap damping triggers on eBGPduplicate
> updates?
> 
> Thanks Bruce for the reply.
> 
> On Sat, 2009-05-09 at 19:50 +0100, Bruce Simpson wrote:
> > Hi Jonathan,
> >
> > Have you actually observed this happening, or is this purely an
> > observation based on code inspection?
> >
> 
> It was based on code inspection.  Duplicating this needed some
> non-trivial test setup (requiring me to replay duplicate bgp updates).
> So I looked at the code, and it looked like it is giving the penalty.
> I was just wondering if someone already looked into this or what was the
> initial design.
> 
> 
> > Jonathan Park wrote:
> > > Hi Xorp users,
> > >
> > > I was looking at XORP code and found that route flap damping can be
> > > triggered by eBGP duplicate updates (i.e. all attributes are the
same),
> > > giving them the same penalty value of 1000.
> > >
> > > Is that true?  If so, is it done intentionally?
> > >
> >
> > It would be really helpful, when reporting this sort of problem on the
> > list, if you could illustrate the call graph(s) involved.
> > I find KScope is an excellent tool for doing this sort of work; it can
> > export call graphs via GraphViz, I believe there may be text export for
> > call trees in 1 direction also.
> >
> > So, 1000 is the FIXED value in damping.cc. Other than that, I haven't
> > had time to read into this further. Can you elaborate further on how you
> > believe this condition is being triggered?
> >
> 
> I think I will do what you suggested and will get back to you with more
> details.
> 
> Thanks,
> 
> Jonathan
> 
> 
> > thanks,
> > BMS
> 
> _______________________________________________
> Xorp-users mailing list
> Xorp-users at xorp.org
> http://mailman.ICSI.Berkeley.EDU/mailman/listinfo/xorp-users