I have successfully established a test network using 3 RB1000 units in a loop configuration using OSPF between each other. Lab2 and Lab3 have a test VRF linked by OSPF to 2 small Soekris boxes running MT4.0b2. All instances of OSPF are set to redistribute connected, static, and bgp; bgp is set to redistribute connected, static, and ospf.

In VPN4 routes, lab2 shows the directly connected subnet with a /30 netmask, but that subnet on lab3 shows as a /32. Lab3 shows the directly connected subnet with a /30 netmask, but that subnet on lab2 shows as a /32...

Is there a bug in the VPN4 route redistribution dealing with /30 networks? The /24 test networks on each ce lab unit are showing up correctly as /24s...

I'll continue to debug, but this is strange behavior and I thought I'd seen a note about this earlier but can't find it right now.

Thanks,

Tim McKee

As I test more, it appears that a redistribute connected in BPG is (on all non-local nodes) showing up in the VPN4 table without a netmask (which I assume is a /32 convention). The local VPN4 table is showing it correctly with a /30 netmask.

As far as I remember this is an old routing-test bug. Try with the latest 3.23 and routing-test.
Be aware there is an OSPF mpls bug in 3.23 that will be fixed in 3.24. It forces you to reboot the pe box before a new ospf instance can be activated.

I guess they haven't fixed it in the 4.0b2 train yet...

Normis, can you comment?

Yes, this is a known bug, which is fixed in v3.23 routing-test.
4.0beta2 is equivalent to v3.22 routing-test, so bug in beta version is still present. Try to downgrading to v3.23

any idea when an updated routing package might be available? I need to be able to test multicast along with MPLS and I don't think you can do that in 3.x.

Resolved and looking good in 4.0b3.