[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: BGPD: crash
- To: misc_(_at_)_openbsd_(_dot_)_org
- Subject: Re: BGPD: crash
- From: Claudio Jeker <cjeker_(_at_)_diehard_(_dot_)_n-r-g_(_dot_)_com>
- Date: Wed, 13 Jul 2005 23:58:30 +0200
- Mail-followup-to: Claudio Jeker <cjeker_(_at_)_diehard_(_dot_)_n-r-g_(_dot_)_com>, misc_(_at_)_openbsd_(_dot_)_org
On Wed, Jul 13, 2005 at 11:22:08PM +0200, Toni Mueller wrote:
> Hi,
>
> I'm running OpenBSD's bgpd on one of my machines (i386) and, for some
> reasons, a snapshot of July 12th. Before, I ran a snap from May 2nd (?)
> which worked very nicely wrt. BGP, with a session lifetime in the
> neighborhood of 10 days (I rebooted the box for unrelated reasons).
> The machine itself has enough horsepower to run bgpd, and I import
> everything.
>
> Now, with the current snapshot, I find this in my logs after bgpd was
> dead:
>
> Jul 13 22:52:08 w1 bgpd[14813]: fatal in RDE: Uh, oh a politician in the decision process
> Jul 13 22:52:08 w1 bgpd[12123]: Lost child: route decision engine exited
>
>
> Any idea on how to (help) debugging this? Who of you runs bgpd
> supervised?
>
For some reasons you end up with two prefixes in the RIB that are
indistinguishable. The decision process should in any case find a more
prefered route or it will fatal with the given message.
Currently I know only one way to produce this by announcing a network
twice, once via config file and the other is added via bgpctl.
Normaly it will "crash" right after you entered the bgpctl command
(bgpctl network add 10.10.0.0/16) and a restart of bgpd will solve the
problem until you reenter the network via bgpctl.
If this is not the case I need your bgpd.conf file and probably some
bgpctl output (for the beginning a "bgpctl show" before the server crashes
would be good).
I will try to fix the problem with announcing a network twice till next
week.
--
:wq Claudio
Visit your host, monkey.org