2.9.16 OSPF redistrubution problem

Mapik · Fri Mar 10, 2006 2:21 pm

I have 3 OSPF areas, with 2.9.11 and routing-test everything works great, but after upgrade to 2.9.16 routers stop redistributing routes between areas
here is config from one router that may redistribute it... Problem persist with both routing and routing-test... router "see" all routes from both areas, but not redistribute. Router is x86 machine. Routers around are RB with 2.9.16 and routing-test. If redistribution router is downgraded to 2.9.11 everything is OK

# mar/10/2006 12:07:56 by RouterOS 2.9.16
# software id = ES7J-JFT
#
/ routing ospf
set router-id=82.150.183.64 distribute-default=never redistribute-connected=as-type-2 redistribute-static=as-type-2 redistribute-rip=no redistribute-bgp=no metric-default=1 \
metric-connected=20 metric-static=20 metric-rip=20 metric-bgp=20
/ routing ospf area
set backbone area-id=0.0.0.0 authentication=none prefix-list-import="" prefix-list-export="" disabled=no
add area-id=172.20.0.0 name="sferia" default-cost=1 stub=no authentication=md5 prefix-list-import="" prefix-list-export="" disabled=no
add area-id=82.150.183.64 name="hamrska" default-cost=1 stub=no authentication=md5 prefix-list-import="" prefix-list-export="" disabled=no
/ routing ospf interface
set FIXME interface=ether1_in cost=1 priority=1 authentication-key="vdYnamicejEsKrytapAternIsIla:\]" retransmit-interval=5s transmit-delay=1s hello-interval=10s dead-interval=2m \
network-type=broadcast
set FIXME interface=ether2_out cost=1 priority=1 authentication-key="vdYnamicejEsKrytasIla:\]" retransmit-interval=5s transmit-delay=1s hello-interval=10s dead-interval=2m \
network-type=broadcast
/ routing ospf network
add network=82.150.183.240/30 area=sferia disabled=no
add network=172.20.65.0/30 area=hamrska disabled=no
/ routing ospf nbma-neighbor
add address=82.150.183.245 poll-interval=1m priority=0 disabled=no
/ routing bgp
set enabled=yes as=1 router-id=0.0.0.0 redistribute-static=no redistribute-connected=no redistribute-rip=no redistribute-ospf=no
/ routing rip
set redistribute-static=yes redistribute-connected=yes redistribute-ospf=no redistribute-bgp=no metric-static=1 metric-connected=1 metric-ospf=1 metric-bgp=1 update-timer=30s \
timeout-timer=3m garbage-timer=2m

PS: Connection tracking SYN times set to 5s

Mon Mar 13, 2006 8:42 am

you have to downgrade only 1 router to 2.9.11 and all others are 2.9.16????

Mapik · Mon Mar 13, 2006 10:45 am

yes, only router that must redistribute between different areas must be 2.9.11 others are 2.9.17 now

Mon Mar 13, 2006 2:00 pm

/ routing ospf 
set router-id=0.0.0.0 distribute-default=if-installed-as-type-1 redistribute-connected=no redistribute-static=no redistribute-rip=no redistribute-bgp=no metric-default=1 metric-connected=20 metric-static=20 metric-rip=20 metric-bgp=20 
/ routing ospf area 
set backbone area-id=0.0.0.0 authentication=none prefix-list-import="" prefix-list-export="" disabled=no 
add area-id=0.0.0.1 name="area1" default-cost=1 stub=no authentication=none prefix-list-import="" prefix-list-export="" disabled=no 
add area-id=0.0.0.2 name="area2" default-cost=1 stub=no authentication=none prefix-list-import="" prefix-list-export="" disabled=no 
/ routing ospf interface 

/ routing ospf network 
add network=192.168.1.0/24 area=backbone disabled=no 
add network=192.168.2.0/24 area=area2 disabled=no 
add network=192.168.10.0/24 area=area1 disabled=no

Three OSPF areas are connected to this router. I am using 2.9.17 routing-test package. works for me!

Mapik · Mon Mar 13, 2006 3:51 pm

I try upgrade to 2.9.17 and problem is still here

Topology is

If blue/green and blue/yellow routers are 2.9.11 and all others are 2.9.17 everything works. In blue area I can see all routes from all networks and in green/yellow area default route is present + routes from this area.

If these two routers are upgraded to 2.9.17 in blue area I see only routes from blue area and in green/yellow area I can see only routes from this area (no default route, no routes from blue area)

Tue Mar 14, 2006 9:30 am

i FOUND A BUG IN YOUR CONFIGURATION!

Where is backbone area???? OSPF can only exchange routes between custom area and backbone area, but not between 2 custom areas!

So instead of area "SFERIA" use default "backbone" (with id 0.0.0.0).

Other solution is something called "virtual links", but i am not familiar with them jet!

Mapik · Tue Mar 14, 2006 10:15 am

Oh, thanks, I will try it

gottin · Tue Mar 14, 2006 10:39 am

I have same problem as Mapik with 2.9.17. My routers are seeing each other , however they don't redistribute their routes. My topology is:
AS1 <=> switch <=> AS2
................... |
............... router

Is it a problem that all of them are in one layer2 segment?

Tue Mar 14, 2006 10:44 am

Create a backbone area!

In this area will be:
router
AS1 border router
AS2 border router

Mapik · Tue Mar 14, 2006 10:47 am

After change all areas have all routes from all areas, really long routing tables

but it's not problem

Thank you wery much

gottin · Tue Mar 14, 2006 10:57 am

Yes, I'm using a backbone area:

/ routing ospf
set router-id=x.y.z.a distribute-default=never redistribute-connected=as-type-2 redistribute-static=no redistribute-rip=no \
    redistribute-bgp=no metric-default=1 metric-connected=20 metric-static=20 metric-rip=20 metric-bgp=20
/ routing ospf area
set backbone area-id=0.0.0.0 authentication=none prefix-list-import=ospf-in prefix-list-export="" disabled=no
/ routing ospf network
add network=212.36.29.16/28 area=backbone disabled=no

I have a rule for setting the distance to 110 in the ospf-in chain

However the state of all the neighbours is continuously changing on all of them - full/2-way/Loading. Is is possible to be a problem of trying to redistribute large number of /32 nets (connected via ppp users)?

gottin · Tue Mar 14, 2006 10:59 am

Mapik, is it a porblem for you to post your successful config, because I still experience problems

Tue Mar 14, 2006 11:19 am

Let as create new aliases:

"AP1 boarder router" will be RB1
"AP2 boarder router" will be RB2
"router" will be GW

so let us proceed:

1) get rid of the all prefix in the area setting

2) create area1 on RB1, assign one network to backbone area, other to area1
3) create area2 on RB2, assign one network to backbone area, other to area2
4) assign inner interface on GW to the backbone area, and set redistribute-default to something!

5) then check and only then try prefix settings

How many route entries are we talking about?

gottin · Tue Mar 14, 2006 12:03 pm

I tried what you offered me.

I removed the prefix-list-import from all areas. All routers 100.100.0.1-3 had ospf network = 100.100.0.0/24 on the backbone area. I also added to additional areas 10.10.0.1-2 that are run on other interfaces as you suggested.

Again there were no distribution of routes and routers were continuously changing there state.

There are about 500 /32 rouite entries on both 100.100.0.2-3. Also 100.100.0.3 is loaded on 100% now (continuously). Even after stopping ospf on other routers. I can not stop ospf on the loaded router. I am starting to think this is not a configuration problem

.

Any ideas what could it be, or how to fix it?

Tue Mar 14, 2006 12:13 pm

Maybe just use /24 or other static routes????? without OSPF it will be much more easier

gottin · Tue Mar 14, 2006 12:40 pm

There are pppoe servers on 100.100.0.2-3 and I is not sure on which one of them a client will connect. That's why a routing protocol is necessary in this situation. I have routed staticly all possible networks.

Tue Mar 14, 2006 1:45 pm

Aha!!! You probably have some kind of routing loop!!!

There are 2 types how two PPPOE servers can coexists on 1 network
1) they have different IP-pool settings
2) they are controled by one RADIUS server

All other configurations can by considared as broken

gottin · Tue Mar 14, 2006 2:00 pm

That is my config. I use one RADIUS server and have different IP-pool settings for all clients with dynamic IPs, however the problem with static routing is with clients that get static IPs. So my routers should understend someway on which access server is connected this client with static address.

It works some way with rip, but sometimes my routers receives wrong information and as result the clients are connecting but there's no correct routes for them

.

Tue Mar 14, 2006 2:16 pm

I don't know!

But if I were you I just use bridge!

gottin · Tue Mar 14, 2006 2:54 pm

What do you mean to use a bridge.

Btw, I'm using routing-test. Should I use on routing package?

Mapik · Wed Mar 15, 2006 10:09 am

I use routing-test too. And for PPPoE users I'm routing everything via OSPF, no problems now. And your configuration is almost the same, it must work

But what can you try is to use /30 networks for connecting routers via OSPF... better results for me.

gottin · Wed Mar 15, 2006 11:34 am

OK, I think I resolved my problem with OSPF. So I will summarise to help others with same problems.

With the topology that I presented above ( 3 routers in one switched network segment ) the solution is:

1) You should have /30 networks between the 3 routers (BR, R1, R2 )
BR 10.8.1.1/30 <=> 10.8.1.2/30 R1
BR 10.8.2.1/30 <=> 10.8.2.2/30 R2
R1 10.8.3.1/30 <=> 10.8.3.2/30 R2

2) in /routing ospf network
add the two neighbouring networks from above on backbone area

3) add a filter for backbone area (prefix-list-import=ospf-in )
/routing filter
1 chain=ospf-in prefix=0.0.0.0/0 prefix-length=32 action=accept

2 chain=ospf-in action=discard

And that's all. This setup works for me - 1 BorderRouter with two AcessServer with PPPoE clients. I hope I wil help to someone with this, because last 3 days my head exploded because of this ...

Wed Mar 15, 2006 12:17 pm

... yes, i forgot about /30

gottin · Wed Mar 15, 2006 9:31 pm

It's working fine without the prefix-list.

The most important is to put the 3 routers to talk to each-other in three different "canals" (/30 nets) instead of one network that takes in an IP address from all the routers.

Btw, I read a lot about OSPF and I think that a configuration that "takes in an IP address from all the routers in one network" shuld not be a problem. However the MT implementation has a small conflict with the official protocol.