OSPF LSDB SIZE reduction

[kinshukpandit]

Hi

I am using two area in my ospf p2p domain. currently my ospf has 45000 lsdb entries. 

How can i reduce my ospf lsdb size.

Thanks in advance 

[M02@rt37]

Hello @kinshukpandit,

You could consider in order to reduce lsdb size:

Implement route summarization at appropriate points in your network. Summarizing routes at the ABR or ASBR can significantly reduce LSDB size. This involves aggregating multiple routes into a single summary route.

Implement LSA filtering to prevent certain types of LSAs from being flooded within an OSPF area. This can help reduce unnecessary LSDB entries.

[kinshukpandit]

Thanks for the reply.

How much lsdb size is recommended for one area. I have LDP running with ospf for mpls L2vpn.

Ldp session are getting down and no ospf flap are observed during this time. i have also used command ldp-sync on my all ospf enabled interfaces.

what can i check in this case.

[M02@rt37]

@kinshukpandit,

The recommended LSDB size for an OSPF area can vary based on factors such as the router's hardware capabilities, memory, CPU resources, and the network's topology. There isn't a strict numerical value for an ideal LSDB size, but keeping it within a manageable range is essential to ensure efficient OSPF operation.

OSPF is designed to handle large networks and databases. It's more about optimizing your OSPF design to avoid excessive LSDB growth.

Do you check LDP adjacencies if they are properly established and stable ?

[kinshukpandit]

yes we check the ldp session they work properly. But some time ldp goes
down abruptly and ospf remain stable on sanme path. This creat issue in
network some time servies remain down for 10 tp 15 min.

[kinshukpandit]

Yes, Generally session for ldp remain stable. but some time ldp goes down without any reason and service remains down for 10 to 15 minutes.

[M02@rt37]

Ok @kinshukpandit,

OSPF LSDB size can impact the overall routing convergence and memory usage of routers. If the LSDB becomes too large, it might lead to increased control plane processing, which could impact other protocols like LDP.

Do you have open a case to Cisco TAC ?

[MHM Cisco World]

I send you private message.

[David Ruess]

Hello,

You can reduce the OSPF LSDB several ways.

Area summarization - reduce Type 3 LSAs

Making an area totally stubby as to allow only the default route type 3 LSA in the are.This will eliminate all LSAs from any other area from coming into the stub area allowing only the type 3 LSA default route.

Prefix suppression - eliminated the prefix information LSA from point to point networks.

Changing the network type between routers. If you have point to point connections on a multi access connection type you can change the link to an OSPF P2P network and this will eliminate the need for a DR/BDR election therefore eliminating the need for the type 2 LSA on that link.

Hope that helps

-David

[Joseph W. Doherty]

@David Ruess, your last paragraph's recommendation didn't occur to me for reducing LSAs, as I always do that to eliminate the needless DR/BDR election, but, yes, it also eliminates a LSA!

Regarding p2p prefix suppression, don't know exactly what you have mind. Might you elaborate?

[David Ruess]

Sure thing @Joseph W. Doherty Funnily enough its its own specific topic in the CCIE exam so it was in my memory bank as an option to reduce LSA's. Now that I think about it and review it may not quire reduce the number of LSA's but it WILL reduce the content in the Type-1 LSAs that's advertised on the links. Cleaning up the DB nonetheless. It will also remove transit links from the routing table since the prefix information is removed. This does mean you won't be able to PING the interface but its a tradeoff as most transit links dont need prefix information for reachability across a network. I'm still learning the nuances of it. 

https://www.cisco.com/c/en/us/support/docs/ip/open-shortest-path-first-ospf/213404-open-shortest-path-first-prefix-suppress.html

It basically removes the prefix information in the LSA 

[Joseph W. Doherty]

@David Ruess, wunderbar!

I wasn't aware of that feature.

I had been wondering if your were implying using unnumbered p2p links.

One of the really nice things about these forums, you can learn much from others.

Thanks for bringing this feature up.

[Giuseppe Larosa]

Hello @kinshukpandit ,

>> I am using two area in my ospf p2p domain. currently my ospf has 45000 lsdb entries. 

Two areas and 45,000 LDSB entries look like that you are using OSPF for customer services prefxes or your redistributing BGP into OSPF.

You should deploy MP BGP and use AF ipv4 unicast for customer services , VPNv4 for MPLS L3 VPN ,other MP BGP AFs can be used for discovery or signalling of MPLS L2 VPN services like VPLS or EVPN. 

You should leave in OSPF only the infrastructure related prefixes links and loopbacks. And you should not deploy redistribute BGP into OSPF for AF IPv4 unicast but deploy MP iBGP everywhere using Route Reflectors.

About LDP sessions instability is hard to say something I wonder why it takes so long 10-15 minutes to recover after a flapping.

If they are LDP targeted sessions between not adjacent routers you need to verify if remote device LDP Router id is present in OSPF link state DB.

Hope to help

Giuseppe

[kinshukpandit]

Bgp redistribution is not present in the network. LDP is used for MPLS l2vpn. All network domain is pure ospf domain. Use of mesh topology make it worse.

[Joseph W. Doherty]

BTW, interestingly, there's another poster with a similar concern, i.e. their is reducing a lab setup with 45,000 LSDB entries.  I.e. https://community.cisco.com/t5/routing/ospf-lsdb-size/td-p/488290