From: Gregg Malcolm (greggm@xxxxxxxxxxxxx)
Date: Wed Apr 03 2002 - 01:04:07 GMT-3
Thanks Parry and to the others that responded. The solution makes sense. I
finally found a copy of Doyle I (I already had II). Been trying to find vol
I for the longest time at my local bookstores. All they ever had was II.
I'm one of those impatient people who can't wait for shipping. I should
have bought it months ago. The section on OSPF admin distance on page 792
was worth $75.00 to me alone. I'm sure I'll have fun trying to read the
whole book tonight.
Thanks again, Gregg
----- Original Message -----
From: "Chua, Parry" <Parry.Chua@compaq.com>
To: "Gregg Malcolm" <greggm@sbcglobal.net>; <ccielab@groupstudy.com>
Sent: Tuesday, April 02, 2002 6:19 PM
Subject: RE: Mutual Redistribution Problem
> When due with multiple points redistribution, don't use filter to block
the route. Change the AD value to control which one should install in
routing table. Use higer metric(cost, hop etc) for redistributed route. This
will prevent loop and sub-optimal route. Redistribute a route learn from
lower AD to a higer AD is not much of issues but not the other way.
>
> Take your topology as reference:
>
> Router R5 and R6 are the routers(RIP/OSPF) that we are concern with.
> In OSPF, when redistribute RIP route into OSPF, it AD value now become 110
and this advertise to all other OSPF domain, in this case, redistributed RIP
route(150.50.30.0) at R5 will be in R6, in R6, the orginal RIP learn
route(via R2) will give way to OSPF due to AD value and prefer the path from
R7 or R5.
>
> To address this problem, at R5 and R6, in OSPF process, set the AD value
of route learn from RIP to 121(higher than RIP). In fact, route
redistributed into OSPF is type Ex, it is easy to change the AD value w/o
route-map and access-list. It is good practice to use both.
>
> Next is the metric use when do redistribution, you should use higher(less
prefer) metric as compare to the orginal route. Take RIP for example, there
are 3 hop across the OSPF domain,
> so we can set metric to 4 so it is less prefence and still reachable.
>
> RIP is a classful routing protocol, the whole 150.50.0.0 will be in RIP
and you may want to filter network that not belong to RIP when redistribute
RIP into OSPF. The other issues is the FLSM and VLSM that need to take care
off.
>
> Parry Chua
>
> -----Original Message-----
> From: Gregg Malcolm [mailto:greggm@sbcglobal.net]
> Sent: Wednesday, April 03, 2002 2:49 AM
> To: ccielab@groupstudy.com
> Subject: Mutual Redistribution Problem
>
>
> Folks,
>
> I'm having trouble with the following scenario. I set it up with the
purpose
> of playing with routing loops.
>
> ______ E0 - 150.50.30.2/27
> |
> R2
> / \
> / \
> / \
> R5 R6
> | |
> | |
> ---------------------
> |
> R7
>
> R2, R5 and R6 are on multipoint frame cloud. They run RIP in the address
> range 150.50.100.0/27.
> R5, R6 and R7 are on Ethernet. They run ospf in the address range
> 150.50.7.0/25
> R5 and R6 do mutual redist of RIP/OSPF.
> I have added a distribute-list on R5 and R6 that denies the 150.50.30
route
> from re-entering OSPF via the Ethernet. Prior to that (as you can
imagine),
> OSPF's lower admin dist caused R5 and R6 to see the subnet as OSPF. I've
> also prevented RIP from re-advertising the .30 subnet from OSPF. (I think
:))
>
> Problem is, when I do a trace from R7 to the .30 subnet, I get sub-optimal
> routing where it bounces from R6 to R5 back to R6 and then to .30. Both
R5
> and R6 show the .30 route as RIP with a next hop of R2. I've played
around
> with DR selection thinking that might be the problem, but no change.
> FLSM/VLSM issues have been addressed. R2 has 2 routes to the R5,R6,R7
> Ethernet and R7 has 2 routes to the .30 subnet.
>
> I'm fairly certain that I could remedy this with policy routing, but I'd
> rather not. I'd like to know why. Any ideas ? This one is kickin my
butt.
>
> Sorry if the ASCII art format is screwed up.
>
> Thanks, Gregg
This archive was generated by hypermail 2.1.4 : Thu Jun 13 2002 - 10:57:52 GMT-3