OSPFv2 — Open Shortest Path First Version 2

Why it exists

Distance-vector protocols exchanged routes and learned about a failure from a neighbour who might themselves be using a looped path. Count-to-infinity was not a corner case; it was the way the information moved.

Link-state inverts the contract. Each router floods a description of its own links, everyone computes the same tree, and a failure is a missing LSA rather than a rumour. OSPF is that idea, specialised for IP, with areas so the database does not have to be the entire AS.

The area is the feature. Without it, every Router-LSA in the AS is in every database, and a flap anywhere runs SPF everywhere.

Down to Full on a broadcast segment

Three routers share a LAN. They elect a DR and a BDR, become fully adjacent only to those two, and flood Router-LSAs so every LSDB matches.

Link
Blocking
Packet in flight
Discarded
Emphasis

Select a device to read its state. Arrow keys walk the topology.

Text equivalent of this diagram

Devices and links at this step
Element	Kind	State
R1 · 1.1.1.1	router	Neighbor state: Down
R2 · 2.2.2.2	router	Neighbor state: Down
Broadcast segment	switch	—
R3 · 3.3.3.3	router	Neighbor state: Down
R1 · 1.1.1.1 — Broadcast segment	link	up · pri 1
R2 · 2.2.2.2 — Broadcast segment	link	up · pri 1
R3 · 3.3.3.3 — Broadcast segment	link	up · pri 1

1 / 8

t = 0 s

Three routers on one Ethernet segment, OSPF just enabled. Every neighbor table is empty. Nothing has been elected, and nothing has been flooded.

What changed

R1 · 1.1.1.1: Neighbor state → Down
R2 · 2.2.2.2: Neighbor state → Down
R3 · 3.3.3.3: Neighbor state → Down
No DR yet

How it works

Neighbors start Down. A Hello moves them to Init; seeing yourself in a Hello moves them to 2-Way. On a point-to-point link that is enough to become adjacent. On a broadcast segment, only the DR and BDR go further with each neighbor; two DROTHERs stay in 2-Way forever, which is correct.

The DR is elected on priority, then Router ID, from the Hellos on the segment. A router that already lists itself as DR is preferred over one that does not, which is why a latecomer with priority 255 does not take over. The BDR exists so that a DR failure does not restart every adjacency on the LAN.

Adjacency is Database Description, then Link State Request, then Full. The Interface MTU lives in the DBD, not the Hello, which is why a jumbo-to-standard pair looks healthy until ExStart and then never leaves it.

The LSDB is typed. Router-LSAs (1) and Network-LSAs (2) stay in the area. Summary-LSAs (3, 4) are what an ABR injects into an area about somewhere else. AS-external-LSAs (5) flood the AS except into stub and NSSA. NSSA uses Type-7 inside the area and translates to Type-5 at the ABR. Opaque 9/10/11 are scoped to link, area, or AS.

SPF runs on the Router- and Network-LSAs of one area, then attaches inter-area and external destinations to that tree. Two routers with the same LSDB compute the same next hops. That is the whole protocol.

A stub area is not a smaller drawing of area 0. It is a filter on flooding. If Type-5 is visible inside it, the area is not stub — or the device is not doing what the configuration claims.

Try it

Elect a DR and BDR

Give R3 a higher priority after R1 is already DR, then switch to a cold start and watch the same numbers produce the opposite result.

When they arrivedAssume a cold start (no Hello claims yet)

Router PriorityR1 1.1.1.1R2 2.2.2.2R3 3.3.3.3

What each Hello currently claimsR1DRBDRR2DRBDRR3DRBDR

DR: R1 · BDR: R2

bdrAmong routers that already list themselves as BDR, R2 wins (priority 1, Router ID 2.2.2.2).
drRouters already listing themselves as DR are preferred, even against a higher-priority latecomer that has not claimed the role. R1 wins (priority 1, Router ID 1.1.1.1).
resultDR is R1, BDR is R2.

Role on this segment
Router	Router ID	Priority	Role
R1	1.1.1.1	1	dr
R2	2.2.2.2	1	bdr
R3	3.3.3.3	255	drother

Run SPF

Change a link cost and step through the candidate list. A router joins the tree when it is the cheapest candidate, and nothing added later can make it cheaper.

Interface costR1 — R2R2 — R3R1 — R4R2 — R5R4 — R5R5 — R6R3 — R6

Iteration 1 of 6

R1 is the root, so it starts on the tree at cost 0.

R2 enters the candidate list at 10, through R1.
R4 enters the candidate list at 10, through R1.

Candidate list after this iteration
Router	Cost	Through
R2	10	R1
R4	10	R1

Resulting routing table at R1
Destination	Cost	Next hop
R2	10	R2
R3	20	R2
R4	10	R4
R5	20	R4
R6	30	R2

Change the area type

Pick stub, NSSA or a Cisco totally-stubby variant and watch which LSA types are actually withheld.

Area typeKind

A stub area refuses AS-external-LSAs. The ABR injects a Type-3 default instead. Type-4 is omitted because it exists only to reach an ASBR.

LSA types 1–5, 7 and opaque 9–11 in a Stub area
Type	Name	Flooded here?	Why
1	Router-LSA	Yes	Router-LSAs describe this area’s routers and stay inside it.
2	Network-LSA	Yes	Network-LSAs describe this area’s multi-access segments and stay inside it.
3	Summary-LSA (network)	Yes	Inter-area prefixes still enter as Type-3, plus a default from the ABR.
4	Summary-LSA (ASBR)	No	Type-4 is not originated into a stub area: there are no Type-5 destinations to reach through an ASBR.
5	AS-external-LSA	No	RFC 2328 section 3.6: AS-external-LSAs are not flooded into stub areas.
7	NSSA-LSA	No	Type-7 is an NSSA mechanism, not a stub-area one.
9	Opaque-LSA (link-local)	Yes	Type-9 is link-local.
10	Opaque-LSA (area-local)	Yes	Type-10 still floods inside the stub area.
11	Opaque-LSA (AS-scope)	No	Type-11 has AS flooding scope, so it is withheld from stub areas the same way Type-5 is.

On the wire

Constructed examples, encoded from the field table below them — not captured traffic.

IPv4: Protocol 89. Destination 224.0.0.5 (AllSPFRouters) on a broadcast segment. RFC 2328
OSPF packet header: Version, type, length, Router ID, Area ID, checksum and authentication. RFC 2328
Hello body: Network mask, HelloInterval, options, priority, RouterDeadInterval, DR and BDR. RFC 2328

Configure it

```
router ospf 1
 router-id 1.1.1.1
```
Process 1 is a local identifier, not an area and not a Router ID. The Router ID is set explicitly so a cloned image cannot elect two boxes as the same originator.
Common mistake: Leaving Router ID to the highest loopback. Cloning a VM then produces duplicate Router-LSAs and an LSDB that cannot describe both devices.
RFC 2328 §12.4.1
```
interface GigabitEthernet0/0
 ip address 192.0.2.1 255.255.255.252
 ip ospf 1 area 0
 ip ospf network point-to-point
```
Puts the link in area 0 and skips DR election. Broadcast is the default on Ethernet and would wait out the Wait Timer for a DR that a p2p link will never need.
Common mistake: Leaving network type at broadcast. Neighbors still form, but they elect a DR and spend 40 seconds doing it.
RFC 2328 §9.1
```
 ip mtu 1500
```
Pins the IP MTU so the Database Description Interface MTU matches the neighbour. This is the field ExStart actually checks.
Common mistake: Setting `ip ospf mtu-ignore` instead of matching MTU. Adjacency comes up; large packets still drop.
RFC 2328 §10.6

Verify

show ip ospf neighbor: The neighbor is Full, network type POINT_TO_POINT, and no DR/BDR is listed.
show ip ospf interface GigabitEthernet0/0: Hello 10, Dead 40, network type POINT_TO_POINT, MTU 1500.

Caveats

IOS-XE default Hello/Dead on broadcast and p2p is 10/40. Mixing a p2p interface with an NBMA default of 30/120 prevents adjacency.

When it breaks

Symptom first, because that is what you have when it happens.

show ip ospf neighbor stuck in EXSTART or EXCHANGE. Hellos increment. Ping works.
Narrow it down
1. Compare IP MTU on both sides of the link, not just Layer-2 MTU.
2. Debug or capture Database Description packets and read Interface MTU.
3. Confirm the network type is what you think — a virtual link sets MTU 0 on purpose.
Cause
The DBD Interface MTU is larger than the receiver will accept, so RFC 2328 section 10.6 discards the packet and master/slave never completes.
Fix
Match the IP MTU. `ip ospf mtu-ignore` only hides the symptom and leaves a forwarding mismatch.
Neighbors never leave Down, or flap every few seconds. The link is up.
Narrow it down
1. Read HelloInterval and RouterDeadInterval in a captured Hello, not in the running-config you think is there.
2. Check both directions. One side at 10/40 and the other at 30/120 is the NBMA/broadcast mix-up.
Cause
HelloInterval or RouterDeadInterval differ, so each side ignores the other’s Hello.
Fix
Make the timers match on the segment. Do not mix broadcast and NBMA defaults on one link.
One side shows a neighbor, the other does not. Or both ignore each other on a link you cabled today.
Narrow it down
1. Compare Area ID in the OSPF header of the Hello.
2. On IOS, check whether the `network` statement or the interface `ip ospf` statement actually placed the link in the area you intended.
Cause
Area ID mismatch. The packet is dropped in RFC 2328 section 8.2 before a neighbor structure is created.
Fix
Put both interfaces in the same area. An ABR is a router in two areas, not a link in two areas.
Adjacency never forms on the ABR-to-stub link. Hellos are visible in a capture.
Narrow it down
1. Compare the E-bit in the Hello Options field.
2. Confirm `area N stub` is present on every router in that area, including the ABR.
Cause
The E-bit must match. A stub router clears it; a normal router sets it. RFC 2328 section 10.5 rejects the Hello.
Fix
Configure stub (or NSSA) on every router in the area. The ABR needs the same statement for that area.
SPF churn, LSAs that bounce, and a Router ID you recognise belonging to the wrong box.
Narrow it down
1. show ip ospf | include ID on every router in the area.
2. Look for a Router-LSA whose advertising router is yours but whose links are not.
Cause
Two routers originated Type-1 with the same Router ID. Sequence numbers fight, and the LSDB cannot describe both sets of links.
Fix
Set Router IDs explicitly. Do not leave them to the highest loopback on a fleet cloned from one image.
A stub area’s LSDB contains Type-5, or it contains no default and no externals.
Narrow it down
1. show ip ospf database on a non-ABR in the area.
2. Confirm the ABR has `area N stub` and is originating 0.0.0.0/0 as Type-3.
Cause
Either the area is not actually stub (so Type-5 is expected) or the ABR is not injecting a default (so the stub has nowhere to send externals).
Fix
Make the area type consistent, and confirm the ABR originates the stub default.

Design notes

Point-to-point on Ethernet links in a spine-leaf underlay skips DR election entirely. Two routers, one adjacency, no Wait Timer, no DROTHER. Broadcast is the wrong network type on a link that will never have a third router.

The default reference bandwidth is 100 Mb/s, so every interface faster than that costs 1. Raising it is a domain-wide change: do it on every router or SPF will prefer a “cheap” gigabit path over a ten-gigabit path that still costs 1.

NSSA exists because a stub area cannot originate externals. A site that must redistribute and also refuse Type-5 from the backbone is an NSSA, not a stub with an exception.

Totally stubby and totally NSSA are Cisco refinements. They suppress Type-3 as well. They are useful; they are not in RFC 2328, and the atlas labels them that way.

Misconceptions

“Traffic goes through the DR.”: The DR is a flooding hub. It originates the Network-LSA and is the adjacency hub on the LAN. Forwarding follows the shortest-path tree, which may not even touch the DR.
“Raising ip ospf priority takes over the DR immediately.”: Not unless the current DR disappears. Section 9.4 prefers routers that already list themselves as DR. Clear the incumbent, or wait for it to die, or clear OSPF on it.
“Areas are just a way to draw the diagram in pieces.”: An area is a flooding domain for Router- and Network-LSAs. Type-5 either enters or it does not, based on area type. If the LSDB in area 1 contains every Type-1 in the AS, you do not have areas; you have labels.
“Every OSPF neighbor should be Full.”: On a broadcast segment, DROTHER–DROTHER stays 2-Way. That is the adjacency graph working. Clearing it to force Full is how people accidentally change network type and elect a DR on a p2p link.

More walkthroughs

Stuck in ExStartfailure

Two routers on a point-to-point link exchange Hellos happily, then hang in ExStart. The Database Description packet carries an MTU the other side will not accept.

Link
Blocking
Packet in flight
Discarded
Emphasis

Select a device to read its state. Arrow keys walk the topology.

Text equivalent of this diagram

Devices and links at this step
Element	Kind	State
R1 · MTU 1500	router	Neighbor: ExStart
R2 · MTU 9000	router	Neighbor: ExStart
R1 · MTU 1500 — R2 · MTU 9000	link	up · point-to-point

1 / 4

t = 0 s

Hellos agree: same area, same mask, same 10/40 timers, same E-bit. The neighbor state moves through Init to ExStart. Operators looking only at Hello counters will say OSPF is fine.

What changed

Hello: R1 · MTU 1500 → R2 · MTU 9000
Hello: R2 · MTU 9000 → R1 · MTU 1500
R1 · MTU 1500: Neighbor → ExStart
R2 · MTU 9000: Neighbor → ExStart
Emphasis on R1 · MTU 1500 ↔ R2 · MTU 9000

A stub area is a flooding boundarydesign-choice

An ASBR originates a Type-5. Area 0 receives it. Area 1 is stub, so the ABR withholds the Type-5 and originates a Type-3 default instead.

Link
Blocking
Packet in flight
Discarded
Emphasis

Select a device to read its state. Arrow keys walk the topology.

Text equivalent of this diagram

Devices and links at this step
Element	Kind	State
ASBR · Type-5	router	—
ABR	router	—
Area 0	zone	—
R3 · stub	router	—
Area 1 stub	zone	—
ASBR · Type-5 — ABR	link	up · area 0
ABR — R3 · stub	link	up · area 1

1 / 3

t = 0 s

The ASBR redistributes a prefix and originates a Type-5 AS-external-LSA. Type-5 has AS flooding scope: it will travel through every non-stub, non-NSSA area.

What changed

Emphasis: Type-5 originator
LSA type 5: ASBR · Type-5 → ABR

Terms

Router-LSA: Type 1. Every router originates one per area, listing its links in that area. In OSPFv2 it also carries stub prefixes; in OSPFv3 it does not.
Designated Router: On an OSPF multi-access segment, the router that originates the Network-LSA and with which every other router becomes fully adjacent. A higher-priority latecomer does not take over.
ExStart: The neighbor state in which two routers elect a master and slave by exchanging empty Database Description packets. Interface MTU is checked here, not in Hello.
Stub area: An OSPF area that does not flood AS-external-LSAs. The ABR injects a Type-3 default instead. Type-4 is omitted because it exists only to reach an ASBR.
LS age: Seconds since the LSA was originated. At 3600 seconds (MaxAge) it is flushed. Flooding prefers the younger of two copies with the same sequence.

Check yourself

R1 is DR with priority 1. R3 boots with priority 255. What happens to the DR on that LAN?
Two DROTHERs on the same LAN. What state should they show toward each other?
Neighbors are stuck in ExStart. Hellos match. What field should you read next?
Which LSA type must not appear in a stub area’s LSDB?
R1 sends HelloInterval 10. R2 sends 30. What happens?
In ExStart, who becomes master of the Database Description exchange?
Why does a stub area not receive Type-4 Summary-LSAs?
Which opaque LSA has AS flooding scope?

Why it exists

Down to Full on a broadcast segment

How it works

Try it

Elect a DR and BDR

Run SPF

Change the area type

On the wire

Configure it

Verify

Caveats

When it breaks

show ip ospf neighbor stuck in EXSTART or EXCHANGE. Hellos increment. Ping works.

Narrow it down

Cause

Fix

Neighbors never leave Down, or flap every few seconds. The link is up.

Narrow it down

Cause

Fix

One side shows a neighbor, the other does not. Or both ignore each other on a link you cabled today.

Narrow it down

Cause

Fix

Adjacency never forms on the ABR-to-stub link. Hellos are visible in a capture.

Narrow it down

Cause

Fix

SPF churn, LSAs that bounce, and a Router ID you recognise belonging to the wrong box.

Narrow it down

Cause

Fix

A stub area’s LSDB contains Type-5, or it contains no default and no externals.

Narrow it down

Cause

Fix

Design notes

Misconceptions

More walkthroughs

Stuck in ExStartfailure

A stub area is a flooding boundarydesign-choice

Terms

Check yourself

Related

OSPFv3

IS-IS

OSPFv2 versus IS-IS