Weird 0.8.11.1 connections spike

[Igor Sysoev]

On Sun, Aug 30, 2009 at 01:45:30AM +1000, Jeff Waugh wrote:

> <quote who="Igor Sysoev">
> 
> > > 2009/08/29 23:15:52 [alert] 8716#0: open socket #152 left in connection 105
> > 
> > This is probably after reconfiguraiton.
> 
> Sorry, yes, I should've checked the time -> that was from when I reverted to
> 0.8.10... there no alerts prior to that.
> 
> > There are signinficant changes in 0.8.11, but you case is strange: I'am
> > testing 0.8.11 on a host with 400r/s with a lot of proxied SSIed
> > subrequests and static files, and there is no connection leak.
> > 
> > Have you built any third party modules ?
> 
> None.
> 
> btw, the majority of my non-static traffic is fastcgi over unix socket. I'm
> going to do the upgrade again and watch the activity.

Could you do the following:

1) enable coredumps
2) set in nginx.conf:
   debug_points  abort;
3) reconfigure nginx, if there are open connections, then nginx creates
   coredump on exit
4) look in log for alerts: open socket #... left in connection NN
5) run "gdb /path/to/nginx /path/to/core", then

   p ((ngx_connection_t *) ngx_cycle->connections[NN]->data)->uri
   p ((ngx_connection_t *) ngx_cycle->connections[NN]->data)->main->count

   do not forget to set NN from log message.


-- 
Igor Sysoev
http://sysoev.ru/en/