From: Roger Dellaca (rdellaca@xxxxxxxxxx)
Date: Mon Dec 18 2000 - 16:09:07 GMT-3
A couple of issues:
1. Your "OSPF" nets appear on your IGRP routers (as long as the mask is the sam
e) in this case because they are out of the same major network. In IGRP (and o
ther DV's), specifying "network 172.16.0.0" does 2 things - 1st, specifies subn
ets to be routed, 2nd, specifies interfaces to send & listen on; passive-int on
ly stops sending, doesn't stop listening or including in routing advertisements
.
If your interfaces on the OSPF side of the redistributing router were a differe
nt major net & that net wasn't included under router igrp, they wouldn't be adv
ertised.
2. Filtering is a good idea, but not always required. In this case, the issue
with IGRP being a lower administrative distance than OSPF is preventing routes
redistributed from OSPF into IGRP from getting back to the redistributing route
r via IGRP. We need to look at more than just the redistributing router.
Let's say you have the following:
SCENARIO A:
R1(s0)------(s0)R2(s1)-----(s0)R3
All links HDLC
OSPF: R1 &R2s0; IGRP: R2s1 & R3
All interfaces are 172.16.x.x/24
No feedback will occur.
BUT take the same config & change as follows:
SCENARIO B:
R1(s0)-----(s0.1)R2(s0.2)-----(s0)R3
All links frame
Feedback will occur if you forget to turn off split-horizon on R3 s0.
Now take the frame scenario, change as follows:
SCENARIO C:
split-horizon left off (by mistake!) & change the addressing to: 10.1.x.x/24 on
the OSPF interfaces, 172.16.x.x/24 on the IGRP interfaces, no network statemen
t for the 10.0.0.0 in the IGRP.
10.0.0.0/8 will be advertised to R3 & will feed back to R2 & will be put in rou
ting table. But when pinging from R3 to 10.1.x.x, when the packet gets to R2 i
t will have the appropriate longer-mask rout ro go to R1.
Now, let's take scenario A and make:
SCENARIO D:
after this works without filtering, you are asked to add an ISDN DDR with float
ing static routes between R2 & R3. You do a manual test of the ISDN by pinging
the ip of the other side's BRI. With the ISDN up, you then get a route loop,
because the OSPF routes that went from R2 to R3 via the serial come back via th
e BRI. Do a distribute-list in under IGRP that either only permits the ISDN ne
ts or denies all the OSPF nets & permits everything else. No interface necessa
ry after "in" because you want it to apply to all interfaces receiving IGRP.
If you add more interfaces to to the IGRP side of R2, let's say R2 to R3 to R4
back to R2 (SCENARIO E?), the same feedback happens & the same type of distribu
te-list in fixes it.
I think that covers feedback, plus I like Doyle's discussion of other probs wit
h redistribution, like his RIP to IGRP with 2 distribution points, using distan
ce commands to solve the probs without losing the fault-tolerance that would be
lost by filtering routes.
>>> "Chuck Larrieu" <chuck@cl.cncdsl.com> 12/16 9:21 PM >>>
These are my first looks into the issue of VLSM and FLSM redistribution. So
please forgive me if this is a bit of a nanve question.
Practice lab scenario - OSPF and IGRP redistribution
Redistribution router
-------------------------
OSPF..............................IGRP
172.16.1.0/24................172.16.24.0/24
172.16.4.64/26...............172.16.23.0/24
Other OSPF routes from elsewhere
----------------------
172.16.2.0/24
172.16.3.0/24
172.16.10.8/30
172.16.10.12/30
172.16.10.4/30
172.16.4.0/26
On the redistribution router, I configure the IGRP process and place the
172.16.0.0 network into it. The bad news is that the IGRP domain sees the
172.16.1.0 network, even though it shouldn't in that it is supposed to be
unique in the OSPF domain. The further bad news is that when I do a one way
redistribution into the OSPF domain, all OSPF routers can ping interfaces of
all routers in the IGRP domain, even though there is no redistribution of
the OSPF routes into the IGRP domain.
So I start fooling with distribute lists, and proceed to make a mess. Routes
disappear. Stop that.
Well, then I do mutual redistribution with no filters. I summarize all my
longer than /24 routes so they will be redistributed into the IGRP domain.
Life is good. I can ping all interfaces of either domain from the other.
But something is nagging at me. What did I miss? Is it true that there are
NO issues in a redistribution scenario if 1) there is only a single point
of redistribution AND 2) summarization at the IGRP (or any classful domain)
occurs at the point of redistribution?
Thanks for letting me think out loud here.
Chuck
----------------------
I am Locutus, a CCIE Lab Proctor. Xx_Brain_dumps_xX are futile. Your life as
it has been is over ( if you hope to pass ) From this time forward, you will
study US!
This archive was generated by hypermail 2.1.4 : Thu Jun 13 2002 - 08:26:05 GMT-3