Hi, today I stumble across a very strange problem. I have a session with a peer who offers me ~204k prefixes. So far, so good. I usually import them into the kernel's routing table by saying 'bgpctl fib couple', but I also have this in my bgpd.conf:
fib-update yes log updates So, in theory, my kernel routing table should also show those ~204k routes + static/ospf/... ones. In the syslog, I can read messages like this when manually saying, in a sequence, $ bgpctl fib decouple $ bgpctl fib couple Jan 26 11:06:20 hostname bgpd[23677]: kernel routing table decoupled Jan 26 11:06:33 hostname bgpd[23677]: kernel routing table coupled But nothing happens to the kernel's routing table. If I say $ bgpctl fib couple two times in a row, only the first one creates such a message in the syslog. The reason why I started playing with this is that out of a sudden, the routing table was decoupled (after working for well over a week), and all announcments to at least my internal peers stopped while I didn't do anything with this box (only fiddled with filters on an internal peer). I checked for communities in bgpd's rib, but nothing. The rib looks fine to me, and ospfd still continues to do it's job on the same box. What I did, though, was working with some static routes that might overlap with some routes inside the BGP rib, in an attempt to go from OSPF announced routes to BGP announced ones. This is on 4.0-stable as of Dec. 18th 2006 on i386. What gives? Best, --Toni++
