On 2009-01-29, Daniel Ouellet <dan...@presscom.net> wrote:
> The reason I asked is because for example I was testing configuration
> using loopback interface and when I reboot and I do not run bgpd I can
> ping the loopback interface no problem, however if I reboot and bgpd run
> I can still ping it, however it will time out regularly and sometime be
> dead for as much as a minutes in worst case and go up/down and the bgp
> sessions with the loopback interface will flap. Not always but sometime
> it does. I just find out by luck I guess when I work doing constant
> ping. if I stop bgpd, all goes normal and no lost packets what so ever,
> if I restart bgpd, then sessions come up, can stay up for a long time no
> problem, but ping time to time to the same loopback will fail and it
> will happened that some bgp sessions will flap. I don't recall have seen
> this on previous version of bgpd and the configuration stayed the same,
> just upgrade to 4.4. It's been running for a few months, but I see rare
> flaps and digging in it, that's what I found.
>
> So, I was curious as to if any loopback interface shouldn't be use with
> bgpd, witch I am pretty darn sure it can be done like any other bgp router.
>
> If I configure the sessions with the interface itself, it's good, if I
> configure sessions with the loopback, I could see time to time flap and
> ping fail to the loopback interface.
I'm using a loopback address on lo1 on my routers, but I assign
the loopback a /32 and distribute that into OSPF.
I don't have a route covering the subnet holding those /32, and
I "route add -reject default 127.0.0.1", so if they're withdrawn
from OSPF the route to the address goes away.
I'm explicitly using these addresses as router-id in both bgpd
and ospfd, and of course running the BGP sessions to (neighbour
address) and from (local-address) those addresses.
A loopback on the local router shows up like this in bgpctl sh fib:
...
flags: * = valid, B = BGP, C = Connected, S = Static
N = BGP Nexthop reachable via this route
r = reject route, b = blackhole route
flags prio destination gateway
* 4 195.95.187.1/32 195.95.187.1
...
and another router's loopback like this:
...
* 32 195.95.187.3/32 195.95.187.243
(this is on -current; on 4.4 you won't have the priorities.
prio 4 is connected, 32 is ospf, bgp would be 48).
