Ospf point to point/point to multi-point

General questions
1

Hi
i am having some issues running ospf network point to point / point to multi-point between a vyos1.2 router and another router (another vyos or cisco). The ospf route exchange successfully, but “show ip route” shows the interface where it connect to its ospf neighbors to be inactive and therefore does not passing the traffic.

show ip route
…
O 192.168.3.0/24 [110/20] via 10.10.1.20, eth0 inactive, 00:03:53
…

this does not happen if i revert to ip network broadcast. it only happens if i set both side to point to point or point to multi-point.

any ideas?

Thanks a lot for your help!

2

to post an update.

the ip ospf network point to point or point to mulitpoint issues occurs reliably on all NBMA (MGRE) tunnels. it doesn’t seem to affect regular tunnel or regular interfaces. the issues is that those route will be learned via ospf database but not injected into the linux routing table (ip route), thus showing the path as inactive . the only way that works is to put the network to ip ospf network broadcast, but sometime that’s isn’t an option.
i suppose this must be a bug?

3

The FRR log should give you an idea what ospfd is doing and why it’s not installing routes from your neighbors.

With your neighbors being over tunnels, are you ensuring that the MTUs are correct with the GRE overhead accounted for?

By default, OSPF won’t complete an adjacency with a neighbor who’s MTU differs from the interface it’s advertising on.
This option can be disabled easily, however it’s good practice to ensure your MTUs match on both ends.

4

thanks for the reply. i will setup the test bed and see the ospfd messages.
i don’t think the MTU are the one to be blamed here. i have disabled the MTU mismatch and it didn’t do a thing. most importantly: it did work with network type broadcast, which implies MTU wasn’t the root cause.

what’s more interesting is that FRR/Quagga implementation did not have this bug. it seems to be unique to VyOS.