Claudio Jeker wrote:
On Mon, Mar 20, 2006 at 02:00:49AM -0500, Daniel Ouellet wrote:
I got bgpd crashing and kill itself in current snapshot of March 18.
Happen twice so far, but I can't see why yet.
Here is the error message I got:
Mar 20 01:34:14 vcnam1 bgpd[18551]: fatal in SE: session_dispatch_imsg:
pipe closed: Operation now in progress
Mar 20 01:34:14 vcnam1 bgpd[20582]: fatal in RDE: pipe write error:
Broken pipe
The error happened before these two lines. You parent process died.
No idea why...
If it is reproducable could you build a bgpd with "make DEBUG=-g" and run
that version in gdb?
I am trying to see if I can reproduce this. So far I can't.
As for any error message, none that I could see in the logs. Just before
from the same process was the session starting:
Mar 20 01:33:10 vcnam1 bgpd[18551]: neighbor x.x.x.x (iBGP to McLean):
state change Idle -> Connect, reason: Start
Mar 20 01:33:13 vcnam1 bgpd[18551]: neighbor x.x.x.x (iBGP to McLean):
state change Connect -> Idle, reason: Stop
Mar 20 01:33:18 vcnam1 bgpd[18551]: neighbor x.x.x.x (iBGP to McLean):
state change Idle -> Connect, reason: Start
Mar 20 01:34:04 vcnam1 bgpd[18551]: neighbor x.x.x.x (iBGP to McLean):
state change Connect -> OpenSent, reason: Connection opened
Mar 20 01:34:04 vcnam1 bgpd[18551]: neighbor x.x.x.x (iBGP to McLean):
state change OpenSent -> OpenConfirm, reason: OPEN message received
Mar 20 01:34:04 vcnam1 bgpd[18551]: neighbor x.x.x.x (iBGP to McLean):
state change OpenConfirm -> Established, reason: KEEPALIVE message received
I even reduce the size and it definitely shouldn't be a problem now:
# bgpctl sh rib memory
RDE memory statistics
180951 IPv4 network entries using 11.0M of memory
214017 prefix entries using 11.4M of memory
40971 BGP path attribute entries using 4.7M of memory
36728 BGP AS-PATH attribute entries using 1.4M of memory,
and holding 40971 references
3509 BGP attributes entries using 137K of memory
and holding 131268 references
3508 BGP attributes using 20.6K of memory
RIB using 28.7M of memory
