On Thu, Oct 22, 2015 at 9:33 AM, Paul Jakma <[email protected]> wrote:

> On Wed, 21 Oct 2015, Daniel Walton wrote:
>
> The following patches change various defaults in quagga. Many of these are
>> timers that were set many years ago (180s BGP holdtime for example) that
>> need to be updated for modern convergence requirements.
>>
>
> Oh, yeah :)
>
> One thing, can be hard to pick the right defaults, and some might want the
> old ones. I wonder could we have some kind of 'profile' command to switch
> between sets of defaults?


We talked about doing something like that internally but in the end decided
it was more complicated than it was worth.  All of these values are
configurable if the user does not like the default. It felt cleaner to have
only one way to configure these.


Daniel Walton (10):
>>  Lower the default MRAI timer for iBGP peers to 0
>>  Lower the default MRAI timer for EBGP peers to zero
>>
>
> The MRAI oh so badly needs to be reduced. I tried having a go at
> persuading IDR on this yonks ago, after reading a couple of interesting
> papers on it. However, without success. It got watered down to:
>
> I'd kind of go for some very low but non-0 value for eBGP though, in case
> there were still others out there with flap-damping enabled.


About ~12 years ago I changed these defaults in IOS, I did 0 for iBGP but
did 5 for EBGP because of the dampening concern.  In hindsight I wish I had
done 0 for eBGP as well.  I cannot remember the last time I saw someone use
dampening...on the off chance that I am peering with someone that is using
it I can always increase the MRAI for that peer.

You can get bgp to converge sub-second (in many scenarios) if these are
zero...that seems worth the trade-off to me.

Daniel
_______________________________________________
Quagga-dev mailing list
[email protected]
https://lists.quagga.net/mailman/listinfo/quagga-dev

Reply via email to