RE: Rip <--->Ospf Redistribution - using distribute-lists?

From: Schulz, Dave (DSchulz@dpsciences.com)
Date: Mon Nov 07 2005 - 12:13:50 GMT-3


Thanks, Chris. So, would you suggest to redistribute (in this case)
based specifically on the routes (using ACLs) and use the distance
command to prefer a specific route/direction?

Dave Schulz,

Email: dschulz@dpsciences.com <mailto:dschulz@dpsciences.com%20>

________________________________

From: Chris Lewis [mailto:chrlewiscsco@yahoo.com]
Sent: Monday, November 07, 2005 8:57 AM
To: Schulz, Dave; kevin gannon; nobody@groupstudy.com; Cisco Nuts
Cc: ccielab@groupstudy.com
Subject: RE: Rip <--->Ospf Redistribution - using distribute-lists?

Tags in either direction would not do the trick, or at least they would
not solve the problem I am describing. For a full description see

http://www.netmasterclass.com/site/member.php?&category=demo&togo=catego
ry=Lib

Click on public files then download the PDF "A scenario with multiple
redistribution points"

The short answer is this.

R2 and R3 both learn 4.4.4.4 via RIP and send it in to OSPF via
redistribution on both routers. All teh taggin you show does is stop
that route from being fed back in to RIP. This does not allow R1 to see
equal cost routes to 4.4.4.4.

The reason is that either R2 or R3 (depending on timing) will become the
ASBR for this route and R1 will only end up seeing one route to 4.4.4.4

The simplest way to solve this is via distance.

The reason I point this out is that the tagging method shown solves a
particular issue, not all issues in redistribution and I do not believe
anyone should use the tag method as a default cure-all method to resolve
mutual redistribution issues.

Chris

"Schulz, Dave" <DSchulz@dpsciences.com> wrote:

        Chris -

        Interesting thoughts. So, by your example, you are saying that
there
        would be mutual redistribution at R2 and R3. Is seems to me that
if you
        use the tag method and both routers, and changed made it
effectively
        doing the reverse at each location, that this should do the
trick. What
        do we believe would happen differently, and, what would be a
better way
        to accomplish this task?

        Dave Schulz,

        Email: dschulz@dpsciences.com

        ________________________________

        From: Chris Lewis [mailto:chrlewiscsco@yahoo.com]
        Sent: Sunday, November 06, 2005 10:29 PM
        To: Schulz, Dave; kevin gannon; nobody@groupstudy.com; Cisco
Nuts
        Cc: ccielab@groupstudy.com
        Subject: RE: Rip <--->Ospf Redistribution - using
distribute-lists?

        Tags as configured below are indeed a nice way to stop routes
from being
        fed back in to their IGP of origin. However, that is often not
enough to
        eradicate sub-optimal routing. For example, take the simple
topology
        below:

        R1--------R2

        | |

        R3--------R4

        Each has a loopback on it from 1.1.1.1 for R1, 2.2.2.2 for R2
etc.

        Now say R2 and R3 both run OSPF and RIP,

        R1 only runs OSPF and R4 only runs RIP.

        The optimal position is to have R1 see equal cost paths to the
loopback
        of R4 via both R2 and R3, however with tags alone, you will not
be able
        to achieve this. It will only see one.

        Netmasterclass has an excellent write up on this in their public
PDF
        section.

        Chris

        "Schulz, Dave" wrote:

        This is a great subject, and one that I have been wrestling with
        for awhile.
        I have used the ACLs for the distribution, but Kevin is
        right...you have to
        maintain the ACLs. And, in a dynamic environment, this may take
        a lot more
        admin than one would care to admit. I have found that easiest
        way to do the
        redistribution is by using tags. I try to keep the same tag
        number as the
        administrative distance to keep things straight. So, if I am
        redistributing
        from OSPF to EIGRP....I use 110 as the tag and 90 in the reverse
        direction.
        This seems to work well and is very easy to implement and keep
        things
        straight.

        The one thing that I have learned with the redistribution is
        applying the
        metrics. For EIGRP, you must apply the metrics either as a
        default in the
        routing process or on the redistribute command line. You cannot
        solely do
        this within the metric. Here is an example of a route-map for
        redistribution
        that I have used:

        router eigrp 1
        redistribute ospf 1 route-map Ospf2Eigrp metric 100000 1 255 1
        1500
        !
        router ospf 1
        redistribute eigrp 1 route-map Eigrp2Ospf metric 10 metric-type
        1
        !
        route-map Eigrp2Ospf deny 10
        match tag 110
        !
        route-map Eigrp2Ospf permit 20
        set tag 90
        !
        route-map Ospf2Eigrp deny 10
        match tag 90
        !
        route-map Ospf2Eigrp permit 20
        set tag 110

        I am interested in any other great ways to accomplish this. You
        can't know
        enough of the ways to do things.

        Dave
        -----Original Message-----
        From: nobody@groupstudy.com
        To: Cisco Nuts
        Cc: ccielab@groupstudy.com
        Sent: 11/5/2005 10:13 AM
        Subject: Re: Rip <--->Ospf Redistribution - using
        distribute-lists?

        Both using distance and or distribute lists is a valid way
        of stopping the loops. The advantage of the distance way
        where possible is you do not have to maintain ACLs so
        as new routes get advertised you do not have to update
        the ACL's.

        It is not always possible to use distance on its own.

        Regards
        Kevin

        On 11/5/05, Cisco Nuts wrote:
> Hello: Is distribute-lists a good idea to use when doing 2-way
> redistribution b/w Rip and Ospf or for that matter b/w any 2
        IGP's in
> general? I have been reading a very good example of this on
        CCO but
        did
> not run into this kind of solution (if I recall correctly)
        when doing
        the
> InetExp Labs. Even a sample lab from DoIt uses the distance
        109 for
        RIP.
> Permitting all RIP routes into Ospf via the distribute-list
        out rip
        under
> Ospf while denying all RIP routes in via the distribute-list
        out ospf
> under Rip seems to nail it down. Any thoughts on this?
        Thanks!!
>
>



This archive was generated by hypermail 2.1.4 : Thu Dec 01 2005 - 09:12:05 GMT-3