Mutual Redistribution Between OSPF and EIGRP at Two Boundary Routers
Two merged companies still run separate routing domains. R1 runs OSPF process 10 (areas 0 and 5) and R2 runs EIGRP AS 100. R1 and R2 are the only redistribution points and both sit in OSPF area 0. R1 has a 1 Gbps link into the EIGRP domain; R2's only path into EIGRP is a 2 Mbps serial circuit. Area 5 attaches to area 0 only through R1, and its routers have no other exit.
Both boundary routers redistribute in both directions. Three problems appeared in the first week. OSPF-derived prefixes entered EIGRP with metric 4294967295 and were never installed by EIGRP routers. After a seed metric of 100000 10 255 1 1500 was added, hosts in area 5 still reach EIGRP destinations through R2 over the 2 Mbps serial instead of through R1 next door; R2 redistributes with metric 10 metric-type 2, while R1 uses defaults. Finally, every time an access link inside the EIGRP domain flaps, area 5 receives withdrawn and re-flooded type-5 LSAs and runs SPF for unrelated prefixes.
1. Why did the OSPF prefixes appear in EIGRP with metric 4294967295, and how should the seed metric be chosen so that redistribution does not distort path selection?
Read the worked answer
EIGRP cannot translate an OSPF cost into its composite metric. The EIGRP metric is built from minimum bandwidth, cumulative delay, load, reliability and MTU. A type-3 or type-5 LSA carries a single cost value and none of those components, so the redistributing router has nothing to fill the metric fields with. With no default-metric and no route-map set metric, IOS advertises the prefix with 4294967295, which EIGRP treats as unreachable and never installs, so downstream routers silently drop the route.
Choosing the seed values is a design decision. Set the bandwidth component to the slowest link a redistributed prefix actually crosses on the OSPF side, not to the 1 Gbps core, otherwise EIGRP routers will believe a prefix behind a 10 Mbps access link has 100 Mbps available and will pull traffic toward it. Delay should be a representative cumulative value in tens of microseconds, and the metric should be deliberately worse than any native EIGRP route for the same destination so redistribution provides reachability rather than attracting transit traffic.
Apply the bulk values with default-metric and use a route-map set metric for exceptions, then verify with show ip eigrp topology that reported bandwidth and delay match the intended worst case.
2. Area 5 hosts egress toward EIGRP destinations through R2 over the 2 Mbps circuit rather than through R1 next door. Explain the mechanism and give two fixes with their trade-offs.
Read the worked answer
OSPF type-2 externals carry the external metric only. The internal cost from the receiving router to the ASBR is not added, so every router in the domain compares a flat 10 (R2) against a flat 20 (R1) and selects R2 regardless of distance. Area 5, whose only exit is R1, sends traffic to R1; R1 forwards it across area 0 to R2 and out the 2 Mbps serial. The result is two extra hops and 50 times less bandwidth than the local exit.
First fix: redistribute as type 1 with metric-type 1. The internal cost to the ASBR is then added, so area 5 routers see a small cost to R1 and stay local, while routers nearer R2 still use R2. This is the correct general answer whenever two ASBRs inject the same external prefixes; the cost is longer SPF computation and external detail that changes when internal topology changes. Second fix: keep E2 but make external metrics reflect topology, for example 20 on R1 and 30 on R2. That works here but optimizes one region at the expense of another, and it must be re-tuned whenever a link changes.
A third option is area 5 filter-list prefix ... in at R1 to block R2's type-5 LSAs from area 5, accepting that area 5 loses a backup path.
3. Every flap of an access link inside the EIGRP domain destabilizes OSPF area 5. What is happening, and how do you contain it without breaking reachability?
Read the worked answer
Each specific EIGRP prefix redistributed into OSPF becomes a type-5 LSA. A link flap withdraws that LSA and re-originates it with a higher sequence number, which is flooded domain-wide and forces an SPF run on every router. Area 5, which has a single exit and no need for per-prefix external detail, absorbs the churn for no benefit.
Contain it in two independent layers. First, summarize at the point of redistribution: on R2, summary-address 10.0.0.0 255.0.0.0 under the OSPF process replaces the component type-5s with one aggregate. An access-link flap then changes nothing as long as at least one component route still matches the summary, which is the largest single win and is independent of area design. Second, contain flooding with area design: make area 5 totally stubby if it has no external paths of its own, or NSSA if it does. Type-5 LSAs are blocked at the ABR and only a default or type-3 is injected.
Validate with show ip ospf database external on an area 5 router before and after, and note that distribute-list inside the OSPF process does not stop LSAs from being originated or flooded; it only filters the local routing table.