Re: EIGRP DUAL Variance

From: Tony Singh <mothafungla_at_gmail.com>
Date: Mon, 7 Oct 2013 15:33:24 +0100

Hi

Lab was on gns3 attempted a few different IOS's and platforms including the VXR 7200's all the same behaviour

I'd like to try it on real hardware time-permitting but hoping someone bails me out

--
BR
Tony
Sent from my iPhone on 3
> On 7 Oct 2013, at 14:29, "Venkataramanaiah.R" <vramanaiah_at_gmail.com> wrote:
> 
> Dear Tony,
> 
> I would suggest you to run a different version of IOS and try out... I
> can only suspect a bug here.
> 
> The behavior you are experience is indeed strange.. It is surprising
> to see the route installed in RIB while it is not shown in topology
> table.
> 
> Regards
> 
>> On 10/7/13, Joe Astorino <joeastorino1982_at_gmail.com> wrote:
>> Hmmmm....also this Cisco white paper seems to agree with the book and with
>> how I thought it worked. They have a simplified version of what you are
>> testing with easier to read numbers.
>> 
>> In the example, the path through D is not chosen because it's metric is not
>> < variance * FD.  However, it also goes on to make this point:
>> 
>> "Also, the reported distance of neighbor D is 25, which is greater than the
>> feasible distance (FD) of 20 through C. This means that, even if variance
>> is set to 3, the E-D-A path is not selected for load balancing because
>> Router D is not a feasible successor."
>> 
>> This seems to certainly be making the case that to be considered for
>> unequal cost load balancing via variance, a path must first pass the
>> feasibility condition.  Clearly, your lab results show otherwise though.
>> 
>> http://www.cisco.com/en/US/tech/tk365/technologies_tech_note09186a008009437d.shtml#var
>> 
>> 
>> 
>> 
>> On Mon, Oct 7, 2013 at 3:34 AM, Joe Astorino
>> <joeastorino1982_at_gmail.com>wrote:
>> 
>>> This certainly seems strange to me.  Granted, it is like 4AM or something
>>> here, but if you had asked me I would have thought EIGRP would not
>>> consider
>>> the F1/0 path for anything, given it doesn't meet the feasibility
>>> condition.  I'd certainly be interested to hear if anybody else has an
>>> answer.
>>> 
>>> 
>>> 
>>> 
>>>> On Sun, Oct 6, 2013 at 2:53 PM, Tony Singh <mothafungla_at_gmail.com> wrote:
>>>> 
>>>> CCNP ROUTE 642-902 Official Certification Guide
>>>> 
>>>> "Routes that are neither successor nor feasible successor can never be
>>>> added to the IP routing table, regardless of the variance setting"
>>>> 
>>>> I did a little lab
>>>> 
>>>>   --- r2 ---
>>>> r1 ----r3 ---- r5 loopback 5.5.5.5
>>>>    ----r4 ----
>>>> 
>>>> 
>>>> no variance
>>>> 
>>>> all-links
>>>> P 5.5.5.5/32, 1 successors, FD is 158720, serno 55
>>>>        via 10.0.0.1 (158720/156160), FastEthernet0/0
>>>>        via 20.0.0.1 (161536/158976), FastEthernet1/1
>>>>        via 30.0.0.1 (1790976/156160), FastEthernet1/0
>>>> 
>>>> topology
>>>> P 5.5.5.5/32, 1 successors, FD is 158720
>>>>        via 10.0.0.1 (158720/156160), FastEthernet0/0
>>>>        via 30.0.0.1 (1790976/156160), FastEthernet1/0
>>>> 
>>>> global-rib
>>>> D        5.5.5.5 [90/158720] via 10.0.0.1, 00:00:44, FastEthernet0/0
>>>> 
>>>> 
>>>> the reason we choose f1/0 path as FS over f1/1 is due to the AD<FD of
>>>> successor holding true even if it has a worse FD - expected behaviour
>>>> and
>>>> end result no load-balancing as no equal cost paths (max 4 equal cost
>>>> paths
>>>> by default)
>>>> 
>>>> 
>>>> 
>>>> 
>>>> if I apply a variance of x2
>>>> 
>>>> 
>>>> 
>>>> all-links
>>>> P 5.5.5.5/32, 2 successors, FD is 158720, serno 52
>>>>        via 10.0.0.1 (158720/156160), FastEthernet0/0
>>>>        via 20.0.0.1 (161536/158976), FastEthernet1/1
>>>>        via 30.0.0.1 (1790976/156160), FastEthernet1/0
>>>> 
>>>> topology
>>>> P 5.5.5.5/32, 2 successors, FD is 158720
>>>>        via 10.0.0.1 (158720/156160), FastEthernet0/0
>>>>        via 30.0.0.1 (1790976/156160), FastEthernet1/0
>>>> 
>>>> global-rib
>>>> D        5.5.5.5  [90/158720] via 10.0.0.1, 00:00:53, FastEthernet0/0
>>>>                  [90/161536] via 20.0.0.1, 00:00:53, FastEthernet1/1
>>>> 
>>>> 
>>>> ok so we're now permitting the f1/1 path being installed as a viable
>>>> load-balancing destination, this shows me two things, first that dual
>>>> will
>>>> permit a non-successor or non-feasible successor into global rib as
>>>> viable
>>>> load-balancing paths and second dual compares the end-to-end FD for
>>>> making
>>>> a decision on which path it would prefer to load-balance looking at the
>>>> all-links topology as f1/1 has a lower FD than f1/0
>>>> 
>>>> 
>>>> 
>>>> this does make the above argument invalid - im wondering if EIGRP has
>>>> since
>>>> been enhanced for this DUAL behaviour? if so then I guess the whole
>>>> point
>>>> for the statement in the first place was to avoid a loop i.e if you're
>>>> not
>>>> a successor or feasible successor then you a loop, what has changed?
>>>> 
>>>> 
>>>> --
>>>> BR
>>>> 
>>>> Tony
>>>> 
>>>> 
>>>> Blogs and organic groups at http://www.ccie.net
>>>> 
>>>> _______________________________________________________________________
>>>> Subscription information may be found at:
>>>> http://www.groupstudy.com/list/CCIELab.html
>>> 
>>> 
>>> --
>>> Regards,
>>> 
>>> Joe Astorino
>>> CCIE #24347
>>> http://astorinonetworks.com
>>> 
>>> "He not busy being born is busy dying" - Dylan
>> 
>> 
>> 
>> --
>> Regards,
>> 
>> Joe Astorino
>> CCIE #24347
>> http://astorinonetworks.com
>> 
>> "He not busy being born is busy dying" - Dylan
>> 
>> 
>> Blogs and organic groups at http://www.ccie.net
>> 
>> _______________________________________________________________________
>> Subscription information may be found at:
>> http://www.groupstudy.com/list/CCIELab.html
Blogs and organic groups at http://www.ccie.net
Received on Mon Oct 07 2013 - 15:33:24 ART

This archive was generated by hypermail 2.2.0 : Fri Nov 01 2013 - 07:35:39 ART