OSPF routes crash

RouterOS Forwarding Protocols
1

I am running RouterOS 6.30.4 with OSPF and MPLS.

If I change something in a router of the network all routes crash in main server. The main server is a CCR 1036.

The cpu goes up and it crashes all the routes. After a minute the cpu goes down and the OSPF starts again.
pppoe, routes, vpls and ospf goes down.


CCR Configuration.

/routing ospf area
add area-id=0.0.0.1 name=clientes
/routing ospf instance
set [ find default=yes ] distribute-default=always-as-type-1 redistribute-connected=as-type-1 router-id=x.x.x.x
/routing ospf area range
add area=clientes range=10.100.0.0/16
add area=clientes range=x.x.x.x/22
/routing ospf interface
add authentication=md5 authentication-key=sss network-type=point-to-point passive=yes
add authentication=md5 authentication-key=sss interface=Loopback network-type=broadcast passive=yes
add authentication=md5 authentication-key=sss interface=ether4 network-type=point-to-point
/routing ospf network
add area=backbone network=172.16.0.0/29
add area=backbone comment=RB2 network=172.16.0.8/29
add area=backbone comment=RB3 network=172.16.0.16/29
add area=backbone comment=LoopBack network=x.x.x.x/32
add area=clientes network=x.x.x.x/22
add area=clientes network=10.100.0.0/16
add area=backbone comment=RB7 network=172.16.0.48/29
add area=backbone comment=“RB 8” network=172.16.0.64/29
add area=backbone comment=“RB 13” network=172.16.0.160/29

Thxs

2

Is this a new problem or did it start after upgrading to a new ROS version?

3

Thanks for reply! I started after increase routers and add a BGP session. I was with RouterOS 6.30 but I upgraded to 6.30.4 to try to solve.

I tried after read the changelog.

What’s new in 6.30.4 (Aug.25.15-12:59:46)

*) route - fixed crash on removing route that was aggregated;


It did not help me to with route crash.

Check up the screenshot. All routing crash at the same time.

4

It only happens at CCR. The other routers are with the same number of routers neighbors does not have it. But the difference is that each route is linked at one interface. The CCR use one ether for all the internal network.

5

Today I have disabled all the Discovery Interface at Neighbor List. After it crashed again! It is something about route table for sure.

6

I’m experiencing the same problem on CCR1016 when disabling ND on interfaces or after removing IP addresses from interfaces - at least since v6.23. Correspondence with MT support did not result in a solution or even acknowledgment of the problem.

7

Have you narrowed down which changes cause a crash? Is it only changes related to OSPF or is it any change in the router?

8

Thanks again! It happens with things that change on route table. It is really strange because crash all routes and increase cpu. Other changes is okay!

9

Now I found out the problem. It is like a loop after a router come off. The ospf goes crazy and cpu really high. Then all the routes falls and comes up after this. I have a summarized area for pppoe clients. I am using ptp as network type on ospf.
I have 6 router in one ether out for a switch. The adjacency stay for days until this type of loop.
How can I solve it? I changed the summarized are to stub and I will try today to use nbma instead.
Thanks

10

Does that mean there is a routing loop in your network. Are you filtering OSPF in any way?

11

Thanks again for your help! No filter in OSPF. Should I use filter for PPPOE clients not propagate? Do I need to filter any more things?
Thanks

12

How many PPPoE sessions do you have and is each one creating an OSPF route?

One thing we had to do when load testing the CCR1072 for PPPoE stability was disable the IPv6 package. Once we did that, we were able to go from 5000 PPPoE sessions to 30,000 PPPoE sessions.

There is some kind of Layer 2 issue in the IPv6 package that limits the performance of IPv4 PPPoE. This may not solve your specific problem, but I would disable IPv6 during the troubleshooting process just to simplify.

13

Thanks again for your help. I am no advertising the pppoe network in OSPF to reduce the interfaces in it.
I am running in this RouterOS BGP peers in the server. Do you think is is the problem to recalculation drop the OSPF?
I am setup a new RouterOS for BGP, firewall and OSPF to connect to authentication server. I will turn the main RouterOS to authentication server only with OSPF with this new serve on the border.

Thanks again.

14

By default max ipv6 neighbors are set to 1024 which may cause trouble with large number of clients. You can try to increase this limit in “/ipv6 settings” menu.

15

Thanks MRZ, we actually increased this value after opening a ticket with MT and set it to 100,000 neighbors but we still did not see significant PPPoE performance gains until we disabled the IPv6 package entirely.

16

Thanks everyone! The problem was BGP with OSPF in the same box. Too many updates of BGP. I solved it!
Thanks for help.