On 2011-01-27, Rod Whitworth <glis...@witworx.com> wrote: > I'm whacked and obviously missing something BUT I did ask a guy who > wrote a seriously good TCP/IP course years ago. > > Scenario: > Datacentre. > 2 Soekris 5501s with a 4 port card = 4 * vr + 4 * sis > One is active bgpd, other is warm standby. > vr0 on #1 is 192.168.1.1 > vr0 on #2 is 192.168.1.2 > There is a cable directly connrcted #1 to #2 > I have been syncing config files over that port<--cable->--port for > ages. > Suddenly it doesn't work. I can't ping either way. ifconfig vr0 on > either shows same media, active, correct IP and netmask., > > Tech went to DC and reports: > Verified no pings work on that network. > Neither can he ping either unit from a PC with an IP in the > 192.168.1.0/24 range. > Changing cable changed nothing else. > tcpdump on each end showed expected aro who-has requests without > answer.when pinging on the tcpdump box. No sign of activity at t'other > end's tcpdump. > pf is not running. > > This is OT because, although we are running OpenBSD and OpenBGPd it > really isn't a problem I could ascribe to the OS. > > OTOH I don't know where there are cluier network people.... > > Help? > A 4*whatever is welcome. :-)
Our vr(4) driver has been known to wedge at times, usually bouncing the interface makes it recover (ifconfig vr0 down; ifconfig vr0 up). http://www.openbsd.org/cgi-bin/cvsweb/src/sys/dev/pci/if_vr.c.diff?r1=1.106;r2=1.107 fixes one problem (and my vr(4) has been stable since then), but PR 6546 suggests that there are other problems which haven't been found yet.