08-05-2013 09:42 AM - edited 03-04-2019 08:39 PM
Hello All,
I have a two routers A and B connected over ethernet link and running OSPF between them. IP ospf network type is configured as point to point. When enabled netflow i see traffic with source as interface IP and destination as 224.0.0.5 consuming more bandwidth. Further diagnosis showed the number of retransmissions of OSPF to be changing at a higher rate. Could anyone let me know the reason for retransmissions.
interface gig0/3
band 4096
ip address x.x.x.x 255.255.255.252
no ip redirects
no ip unreachables
no ip proxy-arp
ip tcp adjust-mss 1372
medias-type rj45
ip ospf network point-to-point
duplex auto
speed auto
Regards,
Sathvik
Solved! Go to Solution.
08-07-2013 10:51 PM
Hi Sathvik,
thanks for the additional information, now we know that the general design seems to follow best practices.
Sent from Cisco Technical Support Android App
08-07-2013 01:02 AM
Hello Sathvik,
could you please tell us a little more about the retransmitted OSPF packets and also about the neighbor-states of your routers?
While exchanging the database, sending the Database Description packets will be retried until the DB exchange is completely done. If this process loops continuously, the problem is often caused by MTU mismatches or other MTU or layer-2 issues.
I noticed that you're doing MSS clamping (ip tcp adjust-mss) with a relativ low value, is the connection type an IPSec-Tunnel or something like that?
Maybe this document helps you in isolating the problem:
http://www.cisco.com/en/US/tech/tk365/technologies_tech_note09186a00800949f7.shtml
Hope that helps
Rolf
08-07-2013 06:39 AM
Hello Rolf,
Thanks for the reply.
Device is in distibuted layer of the production network and there are around 100 access layer devices connected running OSPF,so would not be able to execute debug. command and provide the retransmitted OSPF packet.Neighbours state of the routers is full. MTU throughout the path is set to 1500.
Had a observation from past 2 days, CPU of the device goes high during peak hours so though that might be a issue, but when the CPU utilsation is normal i still see retransmissions. Just a note number of retransmission is not that high its around 2000 for 6 hours.
The document provided is very informative and would help me while troubleshooting OSPF
-sathvik
08-07-2013 06:41 AM
Jusy missed out, IP sec tunneling is used.
Thanks,
Sathvik
08-07-2013 10:58 AM
Hello Sathvik,
and there are around 100 access layer devices connected running OSPF
wow - that's a lot!
How many OSPF devices do you have in a single area?
Recently I read a recommendation about OSPF hierarchy design that said you should avoid fragmentation of Router LSAs. As far as I remember, with a MTU of 1500 byte, a Router LSA can contain at the most 119 Link entries without fragmentation.
Could this be an issue in your network (show ip ospf database)?
Furthermore, very large packets have to be fragmented for the transport through the IPSec tunnel. This shouldn't be a problem but maybe it's worth an examination.
Just a note number of retransmission is not that high its around 2000 for 6 hours.
If the network is stable, we should see "paranoid flooding" every 30 minutes.
2000 : 12 = 166; but this doesn't tell us much. Maybe it helps you in estimating, you know your network.
Best regards
Rolf
08-07-2013 06:10 PM
Fischer is correct. Every 30 minutes, each router, will retransmit LSAs from which it originated. Also, is everything in one big area? If so, if there is any change in Type-1 and Type-2 LSAs, a full SPF run will occur. And since you have a LOT of routers in a single area (not sure on that yet), I could easily see with all the other stuff in your network, an LSA ACK or not coming in on time, or retransmits.
You can also implement, imcremental SPF, LSA Group Pacing, etc for OSPF if you have a lot of routers in one big area. But I would do research on those yourself, and see if it would be a good fit for your network.
Fisher, your post was very informative about how many Type-1 LSas you could contain in a 1500 byte packet.
08-07-2013 10:32 PM
Fischer/Jahn,
OSPF database has 96 router LSA so that should not be an issue. I have 15 area's configured which are totally stub and maximum number of routers in a area is 30. There was a typo error i have 200 Access routers.
Though i am not much aware abt sh ip ospf retransmission-list, executed and observed retransmission happening inbetween.
Thanks,
Sathvik
08-07-2013 10:37 PM
Hi John,
thank you!
After reviewing http://www.itcertnotes.com/2011/12/ospf-type-1-router-lsa.html I have to add that with authentication it's actually less than 119. I also found very much interesting information in "BRKRST-2337 - OSPF Deployment in Modern Networks (2013 Orlando)" - they have really great stuff on Cisco Live 365.
Best reagards
Rolf
08-07-2013 10:51 PM
Hi Sathvik,
thanks for the additional information, now we know that the general design seems to follow best practices.
Sent from Cisco Technical Support Android App
08-09-2013 08:48 PM
Hi Rolf,
Thank you.
Regards,
Sathvik
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide