From: Jamis Buck <jamis@37signals.com>
Date: 2005-11-03T01:17:23+09:00
Subject: Re: [rails] dispatch.fcgi do not exiting ..

This is now fixed in edge rails. TERM will always immediately exit.  
However, it should be noted that shutting down this way can result in  
aborted requests, causing HTTP 500 errors to be served up to the  
client. Thus, the USR1 signals (which previously behaved identically  
to TERM) will still behave "lazily", waiting until after the next  
request to shutdown.

This should allow the listeners to behave better across a restart of  
apache or lighttpd, and still allow USR1 to be used for a clean  
shutdown of just the listeners.

Note that the preferred way of restarting your application is via the  
USR2 signal, which causes the listeners to gracefully restart without  
needing to restart your web server software. However, this approach  
is known to have some issues on some platforms if your listeners are  
managed by the web server itself, so it is really only recommended if  
you are managing your listeners directly (via, for instance, spawn- 
fcgi).

- Jamis

On Nov 2, 2005, at 8:37 AM, Marcin Jurczuk wrote:

> Helo Group.
> I'm trying to setup rails with lighttpd on Linux and have strange  
> problem
>
> When I'm starting lighttpd fcgi processess are spawn OK.
> 16386 ?        S      0:00 /usr/sbin/lighttpd -f /etc/lighttpd/ 
> lighttpd.conf
> 16387 ?        R      0:00 /usr/bin/ruby /var/www/rmgr/public/ 
> dispatch.fcgi
> 16388 ?        R      0:00 /usr/bin/ruby /var/www/rmgr/public/ 
> dispatch.fcgi
> 16389 ?        R      0:00 /usr/bin/ruby /var/www/rmgr/public/ 
> dispatch.fcgi
> 16390 ?        R      0:00 /usr/bin/ruby /var/www/rmgr/public/ 
> dispatch.fcgi
>
> But after shutting it down:
> root@ubuntu:~# rclighttpd stop
> Stopping lighttpd: lighttpd.
> root@ubuntu:~# ps xa|grep ruby
> 16387 ?        S      0:02 /usr/bin/ruby /var/www/rmgr/public/ 
> dispatch.fcgi
> 16388 ?        S      0:02 /usr/bin/ruby /var/www/rmgr/public/ 
> dispatch.fcgi
> 16389 ?        S      0:02 /usr/bin/ruby /var/www/rmgr/public/ 
> dispatch.fcgi
> 16390 ?        S      0:02 /usr/bin/ruby /var/www/rmgr/public/ 
> dispatch.fcgi
> root@ubuntu:~#
>
> dispatch.fcgi processes are still in process table
>
> killing with -TERM signal dosn't help:
> root@ubuntu:~# killall dispatch.fcgi && ps xa|grep dispatch|wc -l
> 5
> root@ubuntu:~# killall -9 dispatch.fcgi
> root@ubuntu:~# ps xa|grep dispatch
> root@ubuntu:~#
>
> This helps :)
>
> I've tried it on Ubunut,SuSE - result is the same..
>
> One thing ...logs showing something like this:
> [02/Nov/2005:16:31:48 :: 16515] asked to terminate ASAP
> [02/Nov/2005:16:31:48 :: 16514] asked to terminate ASAP
> [02/Nov/2005:16:31:48 :: 16513] asked to terminate ASAP
>
> Of course those proces never terminate without my intervention :(
>
> Any ideas ? Any help ?
>
> -- 
> Spock ... Earth ..
>