From: SMTP Test (kellyrg@bp.com)
Date: Thu Nov 27 2003 - 09:49:24 GMT-3
Will you still not redistribute each protocol back into each other even
though the admin distance is higher...I believe the routing protocol will
still redistribute the original-eigrp subnets (learned through OSPF )back
into EIGRP (and vica versa)...as it is the OSPF/EIGRP databases that are
being redistributed...NOT the installed (best route) routing
table........??........This will cause issues.....and more than likely
routing loops, because, as I understand it, this continually recalculating
of the link state database will cause instability....you cannot not
necessarily just rely on admin distances to theorise.
I agree Ahmed, you should avoid multi-point two way redistribution if you
want to GUARANTEE network stability.
-----Original Message-----
From: Weidong Xiao [mailto:Weidong.Xiao@vi.net]
Sent: 27 November 2003 09:40
To: ahmed_mustafa01@excite.com; ccielab@groupstudy.com
Subject: RE: multi redistribution between ospf , eigrp , ripv2
Hi Ahmed,
Thanks for getting back to me.
With respect, I don't agree that routing loop or sub-optimal routing would
definetly happen in my topology.
Let's say network between R1 and R2 is net12, and net56 is the network
between R5 and R6. On R4 under ospf, net12 is imported from eigrp to ospf,
and ospf advertise this route (in form of LSA) to the whole ospf domain.
When this info arrive R3 via ospf domain, R3 will give route to net12 via
ospf domain a distance of 110. At the same time R3 gets route to net12 by
eigrp, which give the route a distance of 90, so route to net12 by eigrp is
prefered.
On R4 under eigrp, net56 is imported from ospf to eigrp, and eigrp advertise
this route to the whole eigrp domain as an external route. When this info
arrives R3 via eigrp domain, R3 will give route to net56 via eigrp domain a
distance of 170. At the same time R3 gets route to net56 by ospf, which give
the route a distance of 110, so route to net56 by ospf is prefered.
This logic can go on and on, and I can't see how routing loop would happen.
I know that loop busters are tweaking of distance and route-filtering by
ACL, tagging, and I can use them pretty fast. I just don't think it's
necessary.
Come on, CCIEs, double triple quintuple CCIEs, Brian, MADMAN, Scott, Peter,
Chuck, Bob,...., come and get me.
-----Original Message-----
From: Ahmed [mailto:ahmed_mustafa01@excite.com]
Sent: 26 November 2003 22:23
To: Weidong Xiao
Subject: RE: multi redistribution between ospf , eigrp , ripv2
Weidoing,
Whenever there are multi-point redistribution, routing loop or sub-optimal
routing would definetly happen. In your scenerio, the actual results can
only be seen by implementing the Scenerio. There are few options, and they
all depend. You could go with access-lists, distance commands, and route
tagging.
Regards,
Ahmed
--- On Wed 11/26, Weidong Xiao < Weidong.Xiao@vi.net > wrote:
From: Weidong Xiao [mailto: Weidong.Xiao@vi.net]
Date: Wed, 26 Nov 2003 13:05:58 -0000
Subject: multi redistribution between ospf , eigrp , ripv2
R1----R2
| |
| | EIGRP
R3----R4 ----------
| | OSPF
| |
R5----R6
Hi group,
In the above topology, R1, R2, R3 and R4 are running EIGRP, R3, R4, R5 and
R6 are running OSPF, R3 and R4 are doing mutual redistribution.
Can some guru confirm that without any route filtering, tagging or twisting
of distance, routing loop or unoptimal routes will not happen?
If this is the case, then if I swap OSPF with RIPv2, or swap EIGRP with
RIPv2 (on R3 and R4 set distance of external ospf to 130), no routing loop
will happen even if I do nothing special?
Thanks,
Weidong
This archive was generated by hypermail 2.1.4 : Fri Dec 12 2003 - 12:29:18 GMT-3