Static Default Beats OSPF Backup on a Dual-Homed Branch
A branch office is dual-homed to two providers through a Nokia 7750 SR. The primary path uses a static default route 0.0.0.0/0 with next-hop 198.51.100.1 and preference 5. The backup path is learned from OSPF on the second uplink as a default route with preference 10. Both uplinks stay up at Layer 1/2. During peak hours the primary provider shows 20–30% packet loss while its interface remains up. VoIP quality suffers, but the router never installs the OSPF backup default. The design has no BFD, no probe, and no route policy. Operations wants failover that reacts to both hard failures and sustained loss, but does not want the branch to flap between providers during brief microbursts. The branch has 300 users, two /24 VLANs, and one 7750 SR with a 1 Gb/s primary and 500 Mb/s backup.
1. Why is the OSPF default not installed, and what would `show router route-table 0.0.0.0/0` show?
Read the worked answer
show router route-table 0.0.0.0/0 will list the static route as active and the OSPF route as inactive or backup, not missing. In SR OS, preference is the tie-breaker for identical prefixes: lower wins. Static default preference is 5; OSPF default is 10. Both are /0, so longest-prefix match is not involved. The static route stays eligible even during 30% loss because it has no reachability check. The OSPF route can be present in the RIB from the protocol, but it is not selected for the FIB. This is expected behavior, not an OSPF failure. The same logic applies to any duplicate prefix: preference, then ECMP if equal.
2. Design a failover that handles both hard failure and sustained loss without flapping. Name the Nokia mechanisms and state the trade-offs.
Read the worked answer
Use BFD on the static route for hard failure, and an SAA/RPM probe with hysteresis for loss. The static route should be withdrawn only after, for example, five consecutive probe failures and restored after ten successes. Keep the OSPF backup at preference 10 so it becomes active when the static route is withdrawn. Do not use 1-second BFD for the loss path; BFD will not see 20–30% loss if packets still arrive. Also apply a route policy to tag the backup and avoid redistribution loops. Trade-off: probes add management traffic and failover is slower than BFD; make timers match the VoIP SLA. Test with a canary branch and verify the return path, because asymmetric routing can still hurt VoIP.