Child Processes are shutting down unexpectedly

[Igor Sysoev]

On Thu, Mar 29, 2007 at 06:38:58PM -0700, Curtis Spencer wrote:

> My nginx seems to hang after a few minutes of no requests.  I am  
> using it to proxy to a single mongrel process in development.
> 
> I am getting the following in my error_log when I am using debug output.
> 
> 2007/03/29 13:18:17 [notice] 30348#0: using the "epoll" event method
> 2007/03/29 13:18:17 [notice] 30348#0: nginx/0.5.15
> 2007/03/29 13:18:17 [notice] 30348#0: OS: Linux 2.6.20-gentoo-r3
> 2007/03/29 13:18:17 [notice] 30348#0: sysctl(KERN_RTSIGMAX): 0
> 2007/03/29 13:18:17 [notice] 30348#0: getrlimit(RLIMIT_NOFILE):  
> 1024:1024
> 2007/03/29 13:18:17 [notice] 30348#0: start worker processes
> 2007/03/29 13:18:17 [notice] 30348#0: start worker process 30353
> 2007/03/29 13:18:17 [notice] 30348#0: start worker process 30354
> 2007/03/29 13:24:29 [notice] 30354#0: signal 28 (SIGWINCH) received,  
> shutting down
> 2007/03/29 13:24:29 [info] 30354#0: epoll_wait() failed (4:  
> Interrupted system call)
> 2007/03/29 13:24:29 [notice] 30354#0: gracefully shutting down
> 2007/03/29 13:24:29 [notice] 30354#0: exiting
> 2007/03/29 13:24:29 [notice] 30353#0: signal 28 (SIGWINCH) received,  
> shutting down
> 2007/03/29 13:24:29 [info] 30353#0: epoll_wait() failed (4:  
> Interrupted system call)
> 2007/03/29 13:24:29 [notice] 30353#0: gracefully shutting down
> 2007/03/29 13:24:29 [notice] 30353#0: exiting
> 2007/03/29 13:24:29 [notice] 30348#0: signal 28 (SIGWINCH) received,  
> stop accepting connections
> 2007/03/29 13:24:29 [notice] 30348#0: signal 17 (SIGCHLD) received
> 2007/03/29 13:24:29 [notice] 30348#0: worker process 30353 exited  
> with code 0
> 2007/03/29 13:24:29 [notice] 30348#0: worker process 30354 exited  
> with code 0
> 2007/03/29 13:24:29 [notice] 30348#0: start worker process 3236
> 2007/03/29 13:24:29 [notice] 30348#0: start worker process 3237
> 2007/03/29 13:24:29 [notice] 30348#0: signal 29 (SIGIO) received
> 2007/03/29 13:24:29 [notice] 3236#0: gracefully shutting down
> 2007/03/29 13:24:29 [notice] 3236#0: exiting
> 2007/03/29 13:24:29 [notice] 30348#0: signal 17 (SIGCHLD) received
> 2007/03/29 13:24:29 [notice] 30348#0: worker process 3236 exited with  
> code 0
> 2007/03/29 13:24:29 [notice] 30348#0: signal 29 (SIGIO) received
> 2007/03/29 13:24:29 [notice] 3237#0: gracefully shutting down
> 2007/03/29 13:24:29 [notice] 3237#0: exiting
> 2007/03/29 13:24:29 [notice] 30348#0: signal 17 (SIGCHLD) received
> 2007/03/29 13:24:29 [notice] 30348#0: worker process 3237 exited with  
> code 0
> 2007/03/29 13:24:32 [notice] 30348#0: signal 28 (SIGWINCH) received,  
> stop accepting connections
> 2007/03/29 13:24:36 [notice] 30348#0: signal 28 (SIGWINCH) received,  
> stop accepting connections
> 2007/03/29 13:24:40 [notice] 30348#0: signal 28 (SIGWINCH) received,  
> stop accepting connections
> 2007/03/29 13:24:42 [notice] 30348#0: signal 28 (SIGWINCH) received,  
> stop accepting connections
> 2007/03/29 13:24:45 [notice] 30348#0: signal 28 (SIGWINCH) received,  
> stop accepting connections
> 2007/03/29 13:25:09 [notice] 30348#0: signal 28 (SIGWINCH) received,  
> stop accepting connections
> 2007/03/29 13:25:52 [notice] 30348#0: signal 28 (SIGWINCH) received,  
> stop accepting connections
> 2007/03/29 13:25:53 [notice] 30348#0: signal 28 (SIGWINCH) received,  
> stop accepting connections
> 2007/03/29 13:26:21 [notice] 30348#0: signal 28 (SIGWINCH) received,  
> stop accepting connections
> 2007/03/29 13:27:08 [notice] 30348#0: signal 28 (SIGWINCH) received,  
> stop accepting connections
> 2007/03/29 13:27:09 [notice] 30348#0: signal 28 (SIGWINCH) received,  
> stop accepting connections
> 2007/03/29 13:27:14 [notice] 30348#0: signal 28 (SIGWINCH) received,  
> stop accepting connections
> 2007/03/29 13:27:23 [notice] 30348#0: signal 28 (SIGWINCH) received,  
> stop accepting connections
> 2007/03/29 13:27:31 [notice] 30348#0: signal 28 (SIGWINCH) received,  
> stop accepting connections
> 2007/03/29 13:28:14 [notice] 30348#0: signal 28 (SIGWINCH) received,  
> stop accepting connections
> 2007/03/29 13:28:15 [notice] 30348#0: signal 28 (SIGWINCH) received,  
> stop accepting connections
> 2007/03/29 13:28:30 [notice] 30348#0: signal 28 (SIGWINCH) received,  
> stop accepting connections
> 2007/03/29 13:29:30 [notice] 30348#0: signal 28 (SIGWINCH) received,  
> stop accepting connections
> 2007/03/29 13:29:32 [notice] 30348#0: signal 28 (SIGWINCH) received,  
> stop accepting connections
> 2007/03/29 13:29:33 [notice] 30348#0: signal 28 (SIGWINCH) received,  
> stop accepting connections
> 2007/03/29 13:29:34 [notice] 30348#0: signal 28 (SIGWINCH) received,  
> stop accepting connections
> 2007/03/29 13:29:36 [notice] 30348#0: signal 28 (SIGWINCH) received,  
> stop accepting connections
> 2007/03/29 13:29:38 [notice] 30348#0: signal 28 (SIGWINCH) received,  
> stop accepting connections
> 2007/03/29 13:29:39 [notice] 30348#0: signal 28 (SIGWINCH) received,  
> stop accepting connections
> 
> Any ideas.  Let me know what additional information would be useful.

Some script in your system sends periodically the SIGWINCH signal
to the master process. This signal says nginx to stop accepting connections.


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