Nginx keep crashes when it is used in uwsgi

[Maxim Dounin]

Hello!

On Thu, Oct 30, 2014 at 10:54:38AM -0400, arulmarans wrote:

> Hi Maxim Dounin,
> 
> I have gone through those steps, still i was not able to catch the backtrace
> in the gdb. 

You may start with providing "nginx -V" output, as recommended at 
http://wiki.nginx.org/Debugging#Asking_for_help.

[...]

> 2014/10/30 13:54:07 [debug] 16259#0: worker cycle
> 2014/10/30 13:54:07 [debug] 16259#0: epoll timer: 74998
> 2014/10/30 13:54:07 [debug] 16259#0: epoll: fd:10 ev:2001
> d:00007F1E6E9F1350
> 2014/10/30 13:54:07 [debug] 16259#0: *2 http keepalive handler
> 2014/10/30 13:54:07 [debug] 16259#0: *2 malloc: 00000000025C8610:1024
> 2014/10/30 13:54:07 [debug] 16259#0: *2 recv: fd:10 0 of 1024
> 2014/10/30 13:54:07 [info] 16259#0: *2 client 172.27.5.166 closed keepalive
> connection
> 2014/10/30 13:54:07 [debug] 16259#0: *2 close http connection: 10
> 2014/10/30 13:54:07 [debug] 16259#0: *2 event timer del: 10: 1414657522682
> 2014/10/30 13:54:07 [debug] 16259#0: *2 reusable connection: 0
> 2014/10/30 13:54:07 [debug] 16259#0: *2 free: 00000000025C8610
> 2014/10/30 13:54:07 [debug] 16259#0: *2 free: 00000000025B2640, unused: 0
> 2014/10/30 13:54:07 [debug] 16259#0: *2 free: 00000000025C8500, unused: 120
> 2014/10/30 13:54:07 [debug] 16259#0: timer delta: 1
> 2014/10/30 13:54:07 [debug] 16259#0: worker cycle
> 2014/10/30 13:54:07 [debug] 16259#0: epoll timer: -1

... here nginx finishes processing of a connection, and goes to 
the kernel, waiting for more connections...

> 2014/10/30 13:54:17 [debug] 16392#0: bind() 0.0.0.0:80 #6
> 2014/10/30 13:54:17 [debug] 16392#0: bind() 0.0.0.0:8090 #7
> 2014/10/30 13:54:17 [notice] 16392#0: using the "epoll" event method
> 2014/10/30 13:54:17 [debug] 16392#0: counter: 00007F9DCDF92080, 1
> 2014/10/30 13:54:17 [notice] 16392#0: nginx/1.7.6
> 2014/10/30 13:54:17 [notice] 16392#0: OS: Linux 2.6.32-358.18.1.el6.x86_64
> 2014/10/30 13:54:17 [notice] 16392#0: getrlimit(RLIMIT_NOFILE): 8192:8192
> 2014/10/30 13:54:17 [debug] 16393#0: write: 8, 00007FFF049D4300, 6, 0
> ===>Restarted (I have monitor proceess whenever it crashes it will be
> restarted)

... and then it's magically restarted, including master process 
which does mostly nothing but monitors worker process - in 
particular, to restart them if needed.

>From the log I suspect that it's your monitor process which does 
the wrong thing.  You may want to try switching it off.

-- 
Maxim Dounin
http://nginx.org/