OSPF Network Types: Broadcast, NBMA and Point-to-Point

Two OSPF routers on a link that is up and passing traffic can still refuse to form an adjacency, and the most common reason is a network-type mismatch. The network type decides whether OSPF floods hellos to a multicast address or unicasts to configured neighbours, whether a DR is elected, and what the hello and dead timers must be. This article breaks down the five types, the timers each inherits, and how to verify which one an interface is really running.

The five types at a glance

• broadcast — default on Ethernet. Multicast hellos to 224.0.0.5, DR/BDR elected, hello 10 s / dead 40 s.
• point-to-point — default on serial and point-to-point subinterfaces. No DR, multicast hellos, hello 10 s / dead 40 s.
• non-broadcast (NBMA) — for Frame Relay and similar multi-access media without multicast support. No DR? No: a DR is elected, but hellos are unicast and every neighbour must be listed with a neighbor statement. Hello 30 s / dead 120 s.
• point-to-multipoint — treats the medium as a collection of point-to-point links. No DR, hellos are multicast and dynamic neighbour discovery works. Hello 30 s / dead 120 s.
• point-to-multipoint non-broadcast — as above but hellos are unicast, requiring static neighbor entries. Used on DMVPN spokes without multicast.

interface GigabitEthernet0/1
 ip ospf network broadcast
!
interface Serial0/0/0
 ip ospf network point-to-point
!
interface Serial0/0/1
 ip ospf network non-broadcast
 ip ospf priority 0
!
interface Tunnel0
 ip ospf network point-to-multipoint

Timers, and why a mismatch kills the adjacency

Hello and dead intervals are part of the hello packet, and OSPF refuses to accept a hello whose timers differ from the local interface. A broadcast interface (10/40) facing an NBMA interface (30/120) will never progress past INIT.

interface GigabitEthernet0/1
 ip ospf hello-interval 30
 ip ospf dead-interval 120
! dead must be 4x hello unless explicitly overridden

If you are forced to interoperate, match the timers explicitly rather than changing the network type on one side only. Note that some platforms support ip ospf dead-interval minimal hello-multiplier 3, but a minimal dead interval cannot be combined with a manually configured dead interval.

DR election and the priority knob

show ip ospf interface GigabitEthernet0/1
!  Network Type BROADCAST, Cost: 1
!  Transmit Delay is 1 sec, State DR, Priority 1
!  Designated Router (ID) 10.0.0.1, Interface address 10.0.1.1
!  Hello due in 00:00:04

interface GigabitEthernet0/1
 ip ospf priority 0        ! never become DR/BDR - typical on access ports

Priority is only consulted at election time, so a router with the highest priority will not preempt an existing DR until the adjacency is reset. Setting priority 0 on access-facing interfaces is a cheap way to keep DR/BDR churn out of your topology.

Symptoms of each mismatch

• Stuck in INIT — the neighbour sees our hellos but we discard theirs: usually a one-way reachability problem on NBMA (missing neighbor statement) or a mismatch in network type that changes the hello destination.
• Stuck in EXSTART/EXCHANGE — adjacency established but database exchange fails: typically an MTU mismatch, not a network-type problem. See OSPF neighbour stuck in INIT for the INIT case.
• Two neighbours on a link but no DR — one side is point-to-point and the other broadcast.
• Neighbour list grows on a tunnel — the interface is broadcast on a medium where only some pairs can hear each other; point-to-multipoint is the correct type.

! the diagnostic sequence
show ip ospf interface brief
show ip ospf neighbor
show ip ospf interface GigabitEthernet0/1 | include Network Type|Hello|Dead
debug ip ospf hello      ! use with care on production gear

NBMA specifics

On a true NBMA interface you must explicitly name every neighbour, because there is no multicast to discover them:

router ospf 1
 neighbor 10.1.1.2 priority 1
 neighbor 10.1.1.3 priority 0

In practice almost nobody runs native NBMA today: point-to-multipoint on a tunnel, or point-to-point subinterfaces, removes the DR election and the manual neighbour list in one step. If you are migrating, the safe order is to change the hub first only when the spokes already run point-to-multipoint, because the hub's change is what breaks DR-dependent spokes.

For how area types interact with these interface types, see OSPF stub and NSSA area types and OSPF LSA types explained.

原文链接：https://study-ccna.com/ospf-network-types/