I have an IBGP full mesh between 8 routers. 4 core routers connected in a ring, and 4 edge routers dual connected to 2 of the core routers, depending which location they are in, 2 to core and 1 and 2 in location A and 2 to core 3 and 4 in location B.
The 2 edge routers in location A have bgp redistribute-internal, it is legacy commands that we may remove but I believe this is causing the problem I cam going to describe below.
I have an EBGP peer on core router 1 at location A, and an EBGP pper on edge router 1 at location A. I turned down the ebgp peer from core router 1 at location A, and right after that, the CPU on both of the edge routers that have the redistribute-internal on spiked to 99, and the OSPF neighbors dropped, probably because the router got overloaded.
Why did this happen, considering that I do not have OSPF on the edge routers saying to redistribute ebgp into it. Does the OSPF process process all of those ibgp routes, but deny them, but still takes the time and CPU to process all 200,000 routes?
I am still unsure why downing an eBGP peer triggered that redistribution, when things were stable.
Other then clear ip bgp * what can trigger that bgp redistribute-internal command to start importing all of the 200,000 ibgp routers it has to ospf?