OSPF flap due to "Too many retransmissions"

Answered Question
Sep 20th, 2008

Hi Team,

We have 4 or 5 routers running on OSPF in MPLS LAN. Out of which only between two routers (one DR-OTHER and one DR) is flapping badly. The logs we are seeing in the router is:

Sep 20 15:52:17.304 IST: %OSPF-5-ADJCHG: Process 10, Nbr x.x.x.x on GigabitEthernet4/1 from FULL to DOWN, Neighbor Down: Too many retransmissions

Sep 20 15:53:17.302 IST: %OSPF-5-ADJCHG: Process 10, Nbr x.x.x.x on GigabitEthernet4/1 from DOWN to DOWN, Neighbor Down: Ignore timer expired

Sep 20 15:53:24.481 IST: %OSPF-5-ADJCHG: Process 10, Nbr x.x.x.x on GigabitEthernet4/1 from LOADING to FULL, Loading Done

Sep 20 16:12:41.810 IST: %OSPF-5-ADJCHG: Process 10, Nbr x.x.x.x on GigabitEthernet4/1 from FULL to DOWN, Neighbor Down: Too many retransmissions

Sep 20 16:13:41.864 IST: %OSPF-5-ADJCHG: Process 10, Nbr x.x.x.x on GigabitEthernet4/1 from DOWN to DOWN, Neighbor Down: Ignore timer expired

Sep 20 16:13:44.580 IST: %OSPF-5-ADJCHG: Process 10, Nbr x.x.x.x on GigabitEthernet4/1 from LOADING to FULL, Loading Done

The logs show the problem is due to "Too many retransmissions".

Kindly help me how do we troubleshoot this OSPF issue due to Too many retransmissions.

thanks in advance.

I have this problem too.
1 vote
Correct Answer by Giuseppe Larosa about 8 years 4 months ago

Hello Arun,

most of the times under a "too many retransmissions" there is an MTU mismatch.

Verify with

sh interface x/y

sh ip inteface x/y

you can use debug ip ospf adj to troubleshoot the issue.

if there's a MTU mismatch you need to fix it with

int x/y

mtu

or you can use

int gx/y

ip ospf mtu-ignore

I would recommend first solution

Database packets have the DF bit set in the ip header. For this reason an MTU mismatch can create problems.

Hope to help

Giuseppe

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 5 (5 ratings)
Loading.
Correct Answer
Giuseppe Larosa Sat, 09/20/2008 - 06:49

Hello Arun,

most of the times under a "too many retransmissions" there is an MTU mismatch.

Verify with

sh interface x/y

sh ip inteface x/y

you can use debug ip ospf adj to troubleshoot the issue.

if there's a MTU mismatch you need to fix it with

int x/y

mtu

or you can use

int gx/y

ip ospf mtu-ignore

I would recommend first solution

Database packets have the DF bit set in the ip header. For this reason an MTU mismatch can create problems.

Hope to help

Giuseppe

arun kumar Sun, 09/21/2008 - 22:14

Hi Giuseppe,

Thanks for your response. By configuring ip ospf mtu-ignore on other routers solved this issue. But I got confused and would like to know one thing.

If there is an MTU mismatch then the adjacency itself will not be formed. Database packets will be exchanged only when the adjacency is formed. And if MTU is different then the OSPF will be stuck in EXSTART state itself (as Database descriptor packet exchange process is elected).

Above is my understanding on OSPF adjacency formation and I would like to know is it correct.

Giuseppe Larosa Mon, 09/22/2008 - 13:42

Hello Arun,

thanks for your kind remarks.

OSPF terminology makes a distinction between neighborship and adjacency.

to be neighbor two routers must see each other in the hello packets, where there is a list of OSPF Router-ID on routers on the link.

This state is the DRother, two routers that are not DR or BDR don't move on.

If one of the two routers is DR or BDR the exstart process is entered and the two routers start to exchange the whole OSPF database.

Actually at first only the LSA headers are exchanged+ lsa sequence number+ lsa age.

At the end of the LSA headers sending the slave machine has built a list of LSA to be requested and makes these requests.

The other router answers with the full info of the requested LSAs.

Only at the end when this pair of routers have their DB in sync they become adjacent, Full adjacent so you see the FULL in the sh ip ospf neigh.

So when you read the OSPF adjacency cannot form with MTU mismatch is meant that when comes the time to exchange the DB the packets having the DF bit set and being as big as possible for efficiency lead to problems.

So a successful DB exchange is required to reach a full adjacency and not the opposite.

Hope to help

Giuseppe

Actions

This Discussion