OSPF Areas and LSA Types, Explained

Introduction¶

OSPF is a link-state protocol: every router builds an identical map of the topology (the link-state database) and runs Dijkstra's algorithm on it to pick shortest paths. That works beautifully until the topology gets large — the database grows, SPF runs get expensive, and a single flapping link churns every router. Areas are how OSPF keeps that in check.

Areas and the backbone¶

An OSPF domain is divided into areas identified by a 32-bit number written like an IP (area 0, area 0.0.0.0, area 51). Rules:

• Area 0 is the backbone. Every other area must connect to it, directly or through a virtual link.
• Routers with all interfaces in one area are internal routers.
• A router with interfaces in two or more areas is an Area Border Router (ABR). ABRs sit on the area 0 boundary and summarise routes between areas.
• A router that injects routes from another protocol (BGP, static, a different OSPF domain) is an Autonomous System Boundary Router (ASBR).

Inside an area, every router has the full link-state database for that area. Across an area boundary, the ABR passes only summary information (prefixes and costs), not the raw topology. So a link flap in area 10 triggers a full SPF in area 10, but only a distance-vector-style update in area 0 and beyond.

The LSA types you'll see¶

Type 6 (multicast OSPF) is effectively dead. Types 1 and 2 describe topology inside an area; 3 and 4 are the ABR telling one area about another; 5 and 7 are about routes learned from outside OSPF entirely.

Type 1 and 2 in practice¶

On a point-to-point link, each router's Type 1 LSA lists the neighbour and cost — no Type 2 needed. On a broadcast segment (Ethernet with 3+ OSPF routers), a Designated Router (DR) is elected; it originates one Type 2 LSA representing the segment, and every router forms a full adjacency only with the DR and BDR. This cuts adjacencies from O(n²) to O(n).

Area types — controlling the flood¶

Stub area variants exist to keep Type 5 externals (which can number in the tens of thousands, e.g. a full BGP table redistributed) out of areas that don't need them:

• Normal area — sees everything: Types 1–5.
• Stub area — no Type 5. The ABR injects a default route (0.0.0.0/0 as a Type 3) instead. No ASBRs allowed inside.
• Totally stubby area (Cisco) — no Type 5 and no Type 3. Only intra-area routes plus a single default. Smallest possible database.
• NSSA (Not-So-Stubby Area) — like stub, but does allow a local ASBR. That ASBR originates Type 7 LSAs, which the ABR translates to Type 5 when flooding into the backbone. Used when a branch has its own external connection.

Every router in an area must agree on the area type, or adjacencies won't form.

Hands-on (Cisco-style config)¶

router ospf 1
 router-id 10.0.0.1
 network 10.1.1.0 0.0.0.255 area 0
 network 10.2.2.0 0.0.0.255 area 10
 area 10 stub
!
! On the ABR, make area 10 totally stubby:
 area 10 stub no-summary

# Verify
show ip ospf neighbor            # adjacencies: state should be FULL (or 2-WAY to non-DR)
show ip ospf database            # the LSDB, grouped by LSA type
show ip ospf border-routers      # known ABRs/ASBRs
show ip route ospf               # O = intra-area, O IA = inter-area, O E2 = external

FRR / Bird / Quagga use similar syntax (area 0.0.0.10 stub in FRR's ospfd).

Verification and troubleshooting¶

• Neighbours stuck in EXSTART/EXCHANGE — MTU mismatch on the link. OSPF checks MTU in DBD packets. Match MTU or set ip ospf mtu-ignore.
• Neighbours stuck in 2-WAY and never go FULL — that's correct between two non-DR routers on a broadcast segment. Only DR/BDR relationships reach FULL.
• Adjacency won't form at all — mismatched hello/dead timers, area number, area type (stub flag), authentication, or subnet mask on the segment. show ip ospf interface shows all of these side by side.
• External routes missing in a stub area — by design. Check the ABR is injecting the default: show ip route 0.0.0.0.
• Suboptimal inter-area path — all inter-area traffic goes via area 0. If two areas are "adjacent" geographically but both only touch the backbone at a distant ABR, traffic hairpins. That's OSPF working as designed; fix it with topology, not config.
• O E1 vs O E2 routes — E2 (the default for redistribution) ignores internal cost to the ASBR; E1 adds it. If two ASBRs advertise the same external prefix and traffic picks the "wrong" one, switch to E1.

Related tools and reading¶

• On-site: Subnet Calculator, IP address lookup, Ping test.
• Related posts: BGP for sysadmins, How NAT actually works.