We have Failover set up over two links, but thanks to ECMP the COST set on the interface is ignored and data flows randomly between the two routers, where both default-routes have a distance of 110, but the OSPF metric is 10 vs 20000. So instead of failover we have “load balancing”. A broken ECMP is to blame, which cannot be turned off.
Both Rap1 and Rap2 routers must use the route via GW1 and HR1. If GW1 fails, the more expensive route via GW2 - HR2 - HR1 should be used, if HR1 fails, the route GW2 - HR2 - backup uplink. Vlan28 have COST 1000, VLAN29 have COST 20000, but they still give 2 default-gw with same distance/cost. Why OSPF ignoring COST settings and send data to both default-gw?
