IGP - redist customer OSPF advertisements

[johnelliot6]

Hi,

We currently run OSPF across our core (7 POPs), to redistribute all static+connected subnets - Works fine when customers only have subnets directly connected, or we statically route additional subnet to them (Our "Internal" OSPF redistributes these subnets to the rest of our POPs)

When a client has two tails back to us(For failover) + an additional subnet(Eg for a firewall behind there router), we typically run OSPF(different process/area Id) on the primary link...customer advertises the additional subnet back to us, and we advertise default route to them, failover link is typically DSL so we have a floating static default route on the CE with higher metric than the OSPF default, and the routed subnet is injected(via radius) into our core with higher metric also than the OSPF route we receive from them.

The issue we have is that the additional subnet they advertise to us, is not redistributed(As expected) into our Internal OSPF(So the rest of our POP's dont know about it), unless we do a "redist ospf N subnets" under our Internal OSPF...we then also need to add a route-map/acl to control what routes are redistributed into our Internal OSPF to ensure customer doesnt potentially black hole us.

Hoping there is a more elegant solution to this...redist ospf process into our Internal ospf just feels like a really bad way to accomplish it?

Cheers for any suggestions.

[Sudeep Valengattil]

Hi John,

For the backup link (DSL), on the core router you can using a static route with higher admin distance (floating route).  So under normal condition, it will be learning the route from the other router which has main interface due to lesser AD.  Once the main interface goes down, OSPF would remove this route and starts using and advertising the floating route.

Regards,

Sudeep

[johnelliot6]

Hi Sudeep,

I think you missunderstood my question - Our current setup works (OSPF/DSL) with additional subnet....I am after a (hopefully) better solution than having to redistribute the clients ospf process(with route-map/acl) into our Internal OSPF....It just doesnt scale well.