[quagga-users 6751] Re: Quagga over CARP Interface

john smith johnsmith0302 at hotmail.com
Thu Apr 13 19:17:26 IST 2006


even with VRRP it would be routing instability, would it not? the routing 
table still needs a change.

more inline


>From: Arnold Nipper <arnold at nipper.de>
>To: john smith <johnsmith0302 at hotmail.com>
>CC: paul at clubi.ie,  Quagga-users at lists.quagga.net
>Subject: Re: [quagga-users 6745] Re: Quagga over CARP Interface
>Date: Thu, 13 Apr 2006 20:10:55 +0200
>
>On 13.04.2006 19:22 john smith wrote
>
>>Why not peer with both and simply set a preference for one over the
>>other? (incoming path prepend/outgoing path prepend) ?
>>
>
>that is BCP but of course introduces instability
>
>>the more the layers of code, the more likely things can break when
>>new features are added.
>>
>
>Of course the feature I'm talking about is nothing we should add to the 
>standard quagga.
>
>iirc Pluris (http://www.lightreading.com/document.asp?doc_id=18297) was 
>working on code taht exactly did that. Moreover they where working on a 
>router which could span up to 70(!!) racks of chassis still behaving like a 
>single box. The main reason for that is simply: stability. Even if you 
>add/drop lines opr part of the box dies you never ever want to have the 
>routing process going down or doing any change because of that,
>

They probably detect on link layer. But still if my bast path was via A and 
now has to shift to B, either i make a linked list of best paths and simply 
swap the 1st one with the second (cuckoo? might as well have 2 links and do 
equal cost if life is so precious?) or recompute, whatver looks better in 
your CPU/hardware benefit.



>
>
>Arnold

_________________________________________________________________
Express yourself instantly with MSN Messenger! Download today it's FREE! 
http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/



More information about the Quagga-users mailing list