Just a status update. I ended up announcing the whole /22 and /23 all the time using the bgp networks statement. Only downside is that if someone filters /23 size prefix and we happen to have the iBGP link down then some incoming traffic gets blackholed. Thanks a lot for quick replies, really helped...

I'd also really like to see GRE-tunnels be possible with Mikrotik. PPTP/L2TP are bit too heavy solutions just to create simple tunnels between Mikrotik and Juniper/Cisco routers.

The issue still remains that with that aggregate it's always active as there will be more specific routes available from the DC2's half of /22.. atleast I can't see how I can specify that create 10.0.1.0/22 aggreate only when there's routes from 10.0.1.0/23 and not from 10.0.3.0/23.

maybe, BGP Aggregate will help? http://wiki.mikrotik.com/wiki/Manual:Routing/BGP#Aggregate use summary-only=yes If I'm reading that documentation correctly it will then supress the more specific advertisiment? So you can't have /22 and /23 that way. For Traffic Engineering(TE) purposes it would be ...

so, if there's 1.1.1.1/32, you want to announce 1.1.0.0/22?.. or some other prefix? you may try to use recursive nexthop lookup... The /32 route will fall inside the /22 prefix, but it's not the nexthop. The idea is as follows: 1.1.1.0/23 is at DC1 1.1.3.0/23 is at DC2 DC1 will announce loopback /3...

Is there a way to do this with a /32 route? So say there's 1.1.1.1/32 via iBGP then announce to eBGP peers the full /22 prefix?

The scenario: 1 AS is shared between two datacenters. The AS has /22 prefix assigned to it, but it's divided internally to two /23s. One per each DC. DCs have one point to point fiber link and each DC has their own transit. Routers are interconnected using iBGP. When that fiber link is up I'd like t...