On Wed, Jun 18, 2008 at 08:51:20PM +0100, Peter Bristow wrote:
> Hi
> I've had a couple of problems with OpenBGPD.
> I have an Extreme Networks switch that has a session to 4.2-RELEASE and
> 4.3-RELEASE OpenBSD route servers.
> 1 of each. Today I saw both OpenBGPD processes crash. Below is the logs from
> the 4.2 server. I haven't got access
> to the 4.3 logs as it crashed hard on bgpd restart and it's in a remote
> location.
>
> Jun 18 17:01:23 rs bgpd[4338]: SE reconfigured
> Jun 18 17:01:23 rs bgpd[12450]: nexthop 92.48.119.135 now valid: via
> 217.112.81.97
> Jun 18 17:01:24 rs bgpd[15659]: RDE reconfigured
> Jun 18 17:02:05 rs bgpd[4338]: neighbor 92.48.95.5: received notification:
> HoldTimer expired, unknown subcode 0
> Jun 18 17:02:05 rs bgpd[4338]: neighbor 92.48.95.5: state change Established
> -> Idle, reason: NOTIFICATION received
> Jun 18 17:02:56 rs bgpd[4338]: Connection attempt from neighbor
> 92.48.95.5while session is in state Idle
> Jun 18 17:03:10 rs bgpd[12450]: rereading config
> Jun 18 17:03:10 rs bgpd[12450]: nexthop 92.48.119.194 now valid: via
> 217.112.81.97
> Jun 18 17:03:10 rs bgpd[12450]: nexthop 92.48.122.137 now valid: via
> 217.112.81.97
>
> Jun 18 17:03:10 rs bgpd[12450]: nexthop 92.48.119.135 now valid: via
> 217.112.81.97 << was the route added prior to the reload.
>
> Jun 18 17:03:10 rs bgpd[12450]: nexthop 92.48.95.2 now valid: via
> 217.112.81.97
> Jun 18 17:03:10 rs bgpd[12450]: nexthop 92.48.95.1 now valid: via
> 217.112.81.101
> Jun 18 17:03:10 rs bgpd[12450]: nexthop 92.48.119.221 now valid: via
> 217.112.81.97
> Jun 18 17:03:10 rs bgpd[4338]: SE reconfigured
> Jun 18 17:03:10 rs bgpd[15659]: RDE reconfigured
> Jun 18 17:03:54 rs bgpd[12450]: rereading config
> Jun 18 17:03:54 rs bgpd[4338]: fatal in SE: session_dispatch_imsg: pipe
> closed: Operation now in progress
> Jun 18 17:03:54 rs bgpd[15659]: fatal in RDE: rde_dispatch_imsg_parent: pipe
> closed
>
> network 92.48.126.128/28 set { nexthop 92.48.119.135 community 64667:0 }
> was the route added and caused the reload.
>
Looks like the parent process died during the 2nd config reload. At least
both the SE and RDE report a closed pipe to the parent. Can you reproduce
the problem or do you have a core file of bgpd lying around?
--
:wq Claudio
