From: vadim shayevich (vshayevich@xxxxxxxxxxx)
Date: Tue Nov 27 2001 - 12:43:56 GMT-3
----- Original Message -----
From: "Ben-Shalom, Omer" <omer.ben-shalom@intel.com>
To: "Dennis #6" <vacant@home.com>; "CCIE Groupstudy"
<ccielab@groupstudy.com>
Sent: Tuesday, November 27, 2001 8:53 AM
Subject: RE: flsm to vlsm
> Sometimes you don't need them and sometimes you do.
> The question to ask is what is faster (in the lab time counts for almost
> anything):
> * Not do the filtering and deal with the consequences if and when they
> appear
> * Automatically apply the filters.
>
> I am not 100% decided on this as well but in general for the distrib
between
> flsm and vlsm I do take this measure.
>
> I personally like the use of tags to deal with this in a generic way IE:
>
> route-map XtoY deny 10
> match tag 1
> route-map XtoY permit 20
> set tag 2
>
> route-map YtoX deny 10
> match tag 2
> route-map YtoX permit 20
> set tag 1
>
> I apply the first to the redist from X to Y and the other to the redist
from
> Y to X, no route feedback is now possible both ways (if you are consistent
> on this even multiple redist from the same X to the same Y are taken care
> of).
>
> This of course only works for X,Y that both supports tags.
>
> when a protocol does not support tags (IE is flsm) I can still guard one
way
> (not redist back to flsm the routes from the flsm) and in this case the
> worry is about the feedback from vlsm back into vlsm, this is mainly a
> problem if the priority of the routes being fed back is higher.
>
> For RIP the admin is higher so route feedback back into vlsm from RIP
> usually is OK (watch the ISDN lines though)
> For IGRP I tend to set the distance to 117 so it is also less favorable
>
> In those cases I just use
> route-map FtoV permit 10
> set tag 3
>
> route-map VtoF deny 10
> match tag 3
> route-map VtoF permit 20
>
> and apply the first to the redist from the flsm to the vlsm and the second
> to the redistribution from the vlsm (that supports tags) into flsm, that
way
> there is not route feedback to the flsm
>
> If the feedback into vlsm of the vlsm own routes is still a problem (ISDN
> cases mainly with OSPF demand circuit) only then do I use distrib lists.
>
> Let me know what you think of this.
>
>
> Thanks
>
> Omer.
>
> -----Original Message-----
> From: Dennis #6 [mailto:vacant@home.com]
> Sent: Tue, November 27, 2001 1:41 PM
> To: CCIE Groupstudy
> Subject: RE: flsm to vlsm
>
>
> Thanks for your thoughts Hansang. While I haven't taken the exam I agree
> with you that there are engineering standards and I would imagine that
> applying these standards will not be penalized in the exam. So my
strategy
> will be to apply engineering standards and good design principles where
ever
> possible.
>
> Dennis
>
>
>
> -----Original Message-----
> From: nobody@groupstudy.com [mailto:nobody@groupstudy.com]On Behalf Of
> Hansang Bae
> Sent: Tuesday, November 27, 2001 2:08 AM
> To: ccielab@groupstudy.com
> Subject: RE: flsm to vlsm
>
>
> At 02:26 PM 11/27/2001 +1100, Albert Lu wrote:
> >[snip]
> >I'm pondering this question as well, as I see from examples in Doyle that
> >redistribution between IGPs should be done carefully, so filtering via
> >route-maps or distribute-list is required. That means you have to use
> >route-maps and distribute-lists for all routers that are doing
> >redistribution between two protocols. But Heck!!! I've seen examples that
> >don't use filtering, and it still works!!
> >My thoughts are that the filtering is there for precautionary measures,
> just
> >in case routes do feed back into the routing protocol it originated from.
> >There's also the admin distance that has to be taken into account of.
> >If anyone has any suggestions of the proper way of doing this, please let
> me
> >know
>
>
> Folks,
>
> While I understand that people are fretting about the smallest detail of
> the lab (e.g. what's consider superfluous command...) there are some
> definite engineering standards that should be used in the real world.
One
> of these is "if you redistribute, you need to route-map/distribute-list
> what your redistributing." If you don't... sooner or later, you'll
> experience rolling blackouts. Just about every book on IGP has examples
> of route feedback causing temporary/rolling blackouts. This won't happen
> every single time as it's a function of topology and routing protocols in
> use. But sooner or later, you will be bitten by this.
> So you should ALWAYS use route-maps (I prefer them to distribute-lists) to
> control redistribution.
>
> hsb
This archive was generated by hypermail 2.1.4 : Fri Jun 21 2002 - 06:45:23 GMT-3