On 2011-04-29, falz <m...@falz.net> wrote:
> I'm setting up some OpenBSD servers to act as routers. I'm setting
> them up as dual homed devices to have BGP running on a loopback
> interface (Lo1) BGP peers will talk loopback to loopback through
> whichever path is valid. OpenOSPFD is used to advertise the loopback's
> /32 into IGP. This is the standard way to do things on routers such as
> Cisco or Juniper.
>
> `ospfctl show rib` shows the loopback IPs, I can ping loopback IPs,
> but BGP will not come up. Troubleshooting with `tcpdump -i lo1` never
> shows any traffic, even my ICMP packets that are definitely from that
> interface. What's going on here? Is the Loopback interface some very
> special device?
Run tcpdump on the interface the packets are received/transmitted on,
not on lo1.
> Is anyone running OpenBGPD between loopbacks on OpenBSD?
Yes. You sometimes get nasty cloned host routes if ospfd bounces,
but this mostly works pretty well. My usual setup is like this:
$ cat /etc/hostname.lo1
inet 192.0.2.5/32
$ grep lo1 /etc/ospfd.conf
interface lo1 { passive }
$ grep 192.0.2.5 /etc/bgpd.conf
router-id 192.0.2.5
local-address 192.0.2.5
