Huge ospf database

[katerina.dardoufa]

Hello!

My company has recently migrated to MPLS. We have two CPEs in the main site an two in the Disaster site. The provider has configured max-lsa 3000 on its PEs connectint to our network. I have included a diagram of the topology with the name of the routers, so I can more easily explain the problem. The ospf neighborship with router Perseas is torn down, because the provider says that every time adjcacency is accomplished, it is torn down because the max-lsa is exceeded. So, we actually only have 3 circuits running.

The command "sh ip ospf database database-summary" on all four routers shows:

Apollon:

Process 1 database summary
  LSA Type      Count    Delete   Maxage
  Router        19       0        0      
  Network       16       0        0      
  Summary Net   2254     0        0      
  Summary ASBR  84       0        0      
  Type-7 Ext    0        0        0      
  Opaque Link   0        0        0      
  Opaque Area   0        0        0      
  Type-5 Ext    2049     0        0      
      Prefixes redistributed in Type-5  5
  Opaque AS     0        0        0      
  Non-self      3014   
  Total         4422     0        0

Artemis:

Process 1 database summary
  LSA Type      Count    Delete   Maxage
  Router        21       0        0      
  Network       19       0        0      
  Summary Net   2213     0        0      
  Summary ASBR  62       0        0      
  Type-7 Ext    0        0        0      
  Opaque Link   0        0        0      
  Opaque Area   0        0        0      
  Type-5 Ext    2049     0        0      
      Prefixes redistributed in Type-5  4
  Opaque AS     0        0        0      
  Non-self      3424   
  Total         4364     0        0   

Kassiopi:
Process 1 database summary
  LSA Type      Count    Delete   Maxage
  Router        17       0        0      
  Network       17       0        0      
  Summary Net   1400     0        0      
  Summary ASBR  45       0        0      
  Type-7 Ext    0        0        0      
  Opaque Link   0        0        0      
  Opaque Area   0        0        0      
  Type-5 Ext    2049     0        0      
      Prefixes redistributed in Type-5  0
  Opaque AS     0        0        0      
  Non-self      3058   
  Total         3528     0        0  

Perseas:

Process 1 database summary
  LSA Type      Count    Delete   Maxage
  Router        15       0        0      
  Network       16       0        0      
  Summary Net   956      0        0      
  Summary ASBR  24       0        0      
  Type-7 Ext    0        0        0      
  Opaque Link   0        0        0      
  Opaque Area   0        0        0      
  Type-5 Ext    2049     0        0      
      Prefixes redistributed in Type-5  0
  Opaque AS     0        0        0      
  Non-self      3058   
  Total         3060     0        0  

Questions:

1. If the provider has max-lsa 3000, but the ospf database on all our routers is greater than 3000, why doesn't he tear down all adjacencies? I think that the max-lsa has to do with advertisments incoming to the PEs, that is originating from our network.

2. How can I reduce the size of the database, so that everyone is happy, and everything is still working?

Thanks in advance,

Katerina

[Giuseppe Larosa]

Hello Katerina,

you have 2049 external routes and this should be investigated

are you using redistribute connected extensively in your network ?

you should move to network commands + passive interface in order to minimize the number of external routes

Your routers can only flood external routes but at area boundary inter area routes can be filtered or summarized and this is what you can do to reduce the number of LSA.

About max-lsa used on PE nodes

http://www.cisco.com/en/US/docs/ios/iproute_ospf/command/reference/iro_osp2.html#wp1012804

the number is that of non locally generated LSAs so the perception of the PE is different in different sites:

when a PE generates an OSPF LSA from a MP BGP VPNv4 route this does not count for the max-lsa limit

on the site where most of the routes are learned from CE max lsa is more effective.

Hope to help

Giuseppe

[katerina.dardoufa]

Hello Guiseppe and thank you for your reply!

It is interesting that we have so many external routes! On one of the routers in the topology (ABR) I looked at "show ip ospf database external" (I don't know if this is the correct way to approach the problem - I only new OSPF in theory, now I am diving deeper) and I can see that most of the info comes from the provider (about 2000 entries!!!)

Since we are connected through OSPF area 0 with the provider, it only seems natural to me, that in order to have the same topology we will flood all this info back to him (maybe I am missing something here!)

Since most of the routes are coming from the provider, shouldn't he try to somehow summarize and filter what he is sending, so I don't send it back???

Thanks in advance,

Katerina