OPERATOR REFERENCE · OSPFv2

OSPF Field Reference
Cisco IOS-XE & Arista EOS

verify → inspect → configure → tune → troubleshoot — one page, two vendors

Scope note: covers OSPFv2 (IPv4) on Cisco IOS-XE and Arista EOS — both track RFC 2328 closely and their CLI is near-identical for OSPF. OSPFv3 (IPv6) uses the same state machine but different LSA numbering — not covered here. Default timers assume no manual override.

ROUTER-A · Area 0

OSPF

ROUTER-B · Area 0

01 Verify & Inspect

✓

Neighbor Status

Cisco IOS-XE show ip ospf neighbor show ip ospf neighbor detail

Arista EOS show ip ospf neighbor

State column should read FULL (or 2WAY on a broadcast segment for a DROTHER-DROTHER pair, which is normal, not a fault).

▤

Link-State Database

Cisco / Arista show ip ospf database show ip ospf database router

Every router in an area must have an identical LSDB for that area — mismatched sequence numbers point at a stuck adjacency, not just a slow one.

◎

Interface State

Cisco / Arista show ip ospf interface show ip ospf interface brief

Area, process ID, cost
Network type (broadcast/P2P/NBMA)
DR / BDR, priority
Hello / dead timers

→

Routes & Process Summary

Cisco / Arista show ip route ospf show ip protocols

show ip protocols confirms which interfaces are actually participating vs. passive.

⏱

Timers Check

show ip ospf interface Gi0/1

Hello/dead intervals must match exactly between neighbors on the same segment — a mismatch prevents adjacency from forming at all.

🔒

Authentication

show ip ospf interface Gi0/1 | include authentication

Verify auth type (null/plaintext/MD5) matches on both ends, not just that a key is configured.

⚑

First-Look Diagnostics

ping x.x.x.x
show ip ospf
show ip ospf border-routers
debug ip ospf adj (Cisco, use sparingly)
debug ip ospf hello

◍

Redistribution Check

show ip ospf — look for "Redistributing External Routes"
External routes show as O E1 (cost = internal + external) or O E2 (cost = external only, default)

02 Base Configuration

Enable Process

Cisco / Arista router ospf 1 router-id 1.1.1.1

Process ID is locally significant only — doesn't need to match between routers.

net

Network Statement (classic)

network 10.0.0.0 0.0.0.255 area 0

Wildcard mask, not a subnet mask — inverse of a normal mask (0.0.0.255 = /24).

Interface-Level (modern)

interface Gi0/1 ip ospf 1 area 0

Preferred on current IOS-XE/EOS — avoids wildcard-mask ambiguity, explicit per interface.

🔇

Passive Interface

router ospf 1 passive-interface default no passive-interface Gi0/1

Suppresses hellos everywhere except explicitly excluded links — safer default than opting in per-interface.

⚖

Interface Cost

ip ospf cost 10

Overrides the auto-calculated cost (reference-bandwidth ÷ interface bandwidth) — see section 06.

👑

DR Priority

ip ospf priority 100

Default = 1. Priority 0 means "never DR/BDR" on that interface.

🔒

MD5 Authentication

ip ospf authentication message-digest ip ospf message-digest-key 1 md5 MyKey

Key ID and key string must match on both ends of the segment.

↑

Default Route Injection

default-information originate

Add always to advertise 0.0.0.0/0 even without a local default route.

✎

Reference Bandwidth

auto-cost reference-bandwidth 100000

⚠ must be set identically on every router in the domain, or cost comparisons become inconsistent

03 Neighbor Adjacency State Machine

DOWN

no hellos seen

→

INIT

hello received, own RID not seen yet

→

2-WAY

bidirectional, DR/BDR elected here

→

EXSTART

master/slave negotiated

→

EXCHANGE

DBD packets swapped

→

LSRs sent for missing LSAs

→

FULL

LSDBs synchronized

On broadcast/NBMA segments, DROTHER-to-DROTHER neighbors intentionally stop at 2-WAY — they don't need full adjacency with each other, only with the DR/BDR. That's expected, not stuck. Attempt (NBMA-only, unicast poll before Init) is omitted above as an edge case.

04 DR / BDR Election

Why It Exists

On a multi-access segment, every router forming a full mesh of adjacencies would mean N×(N-1)/2 flooding relationships. The DR (and backup BDR) becomes the single point every other router adjacencies with — flooding scales linearly instead.

Election Rule

Highest priority wins; tie broken by highest router ID. Priority 0 removes a router from eligibility entirely.

Non-Preemptive

⚠ election is non-preemptive — a router with a higher priority joining later does not take over an already-elected DR. It only takes effect after the current DR is removed (or the segment resets), which surprises people expecting live failover on config change.

Only on Multi-Access

DR/BDR election happens on broadcast and NBMA network types only. Point-to-point and point-to-multipoint links skip it entirely — see section 06.

Confirm Current DR

show ip ospf interface Gi0/1

Shows DR/BDR IP, plus this router's own state relative to them (DR, BDR, or DROTHER).

Force Re-Election

clear ip ospf process

⚠ resets every adjacency on the box, not just one interface — disruptive, use in a maintenance window

05 Area Types & LSA Types

Area Types

Type	Allows
Backbone (Area 0)	Everything — all other areas must connect through it (or a virtual link)
Standard	Intra-area, inter-area, and external (Type 5) routes
Stub	Blocks Type 5 externals — injects a default route instead
Totally Stubby	Blocks Type 5 and Type 3 summaries — default route only (Cisco proprietary term, widely supported)
NSSA	Like Stub, but allows local Type 7 externals (translated to Type 5 at the ABR)
Totally NSSA	NSSA + blocks Type 3 summaries too

Common LSA Types

Type	Name	Originated By	Scope
1	Router LSA	Every router	Area-local
2	Network LSA	DR on multi-access segments	Area-local
3	Summary LSA	ABR	Inter-area
4	ASBR Summary LSA	ABR	Advertises reachability to the ASBR
5	AS External LSA	ASBR	Domain-wide (not into stub areas)
7	NSSA External LSA	ASBR inside an NSSA	NSSA-local, translated to Type 5 at ABR

06 Network Types, Cost & Timers

Network Types

Type	DR/BDR?	Default Hello / Dead
Broadcast (Ethernet)	Yes	10s / 40s
Point-to-Point	No	10s / 40s
NBMA (e.g. Frame Relay)	Yes	30s / 120s
Point-to-Multipoint	No	30s / 120s

Dead interval is 4× hello by default on every type — override one and the other doesn't follow automatically.

Cost Calculation

Cost = reference-bandwidth ÷ interface bandwidth (both in the same units, result truncated to an integer, minimum 1).

Default reference-bandwidth is 100 Mbps (10^8) — a legacy default from when Fast Ethernet was fast. On modern networks with 1G/10G/40G+ links, every interface ≥100 Mbps computes to cost 1, making OSPF unable to distinguish a 1G link from a 40G link by cost alone.

VERIFY — bump auto-cost reference-bandwidth domain-wide (e.g. to 100000 for 100G-normalized costs) on greenfield builds; changing it on a live network requires updating every router or path costs become inconsistent mid-migration.

07 Why Adjacency Won't Form

Stuck in	Likely cause
DOWN	No hellos arriving — link down, ACL blocking IP proto 89, wrong interface config
INIT	One-way hello — check for asymmetric routing or a unidirectional link fault
2-WAY	Expected for DROTHER-DROTHER pairs on multi-access; otherwise check area ID, subnet mask, or MTU mismatch blocking DBD progress
EXSTART	MTU mismatch is the classic cause — DBD packets are sent at interface MTU and silently dropped if they don't match
EXCHANGE / LOADING	LSA request/response not completing — often a flaky link dropping large packets intermittently
FULL but no routes	Route filtering, distribute-list, or the prefix simply isn't in the LSDB — check show ip ospf database

Fast Checklist

Area ID matches on both sides of the link
Subnet mask matches (or interface-mode config used instead)
Hello/dead timers match
Authentication type + key match
MTU matches — the #1 cause of an EXSTART/EXCHANGE stall
Interface not accidentally marked passive
No ACL blocking IP protocol 89 (OSPF has no L4 port — it rides directly on IP)

OSPF uses multicast 224.0.0.5 (all OSPF routers) and 224.0.0.6 (all DR routers) — confirm multicast isn't being filtered on the segment.