Upgrading binaries with USR2 results in two severs running

Shouldent the original process die after it starts the new process? I
see another server start up, but the old one hang around:

Log:
2010/11/03 07:16:41 [notice] 54825#0: signal 12 (SIGUSR2) received,
changing binary
2010/11/03 07:16:41 [notice] 54825#0: changing binary
2010/11/03 07:16:41 [notice] 54825#0: start new binary process 54951
2010/11/03 07:16:41 [notice] 54951#0: using the “epoll” event method
2010/11/03 07:16:41 [notice] 54951#0: getrlimit(RLIMIT_NOFILE):
1024:1024
2010/11/03 07:16:41 [notice] 54951#0: start worker processes
2010/11/03 07:16:41 [notice] 54951#0: start worker process 54952
2010/11/03 07:16:41 [notice] 54951#0: start worker process 54953

Before:

ps -ef | grep nginx

root 54825 1 0 18:09 ? 00:00:00 nginx: master process
/opt/local/contrib/sbin/nginx -p /var/log/nginx
www-data 54932 54825 0 18:15 ? 00:00:00 nginx: worker process

www-data 54933 54825 0 18:15 ? 00:00:00 nginx: worker process

After:

ps -ef | grep nginx

root 54825 1 0 18:09 ? 00:00:00 nginx: master process
/opt/local/contrib/sbin/nginx -p /var/log/nginx
www-data 54932 54825 0 18:15 ? 00:00:00 nginx: worker process

www-data 54933 54825 0 18:15 ? 00:00:00 nginx: worker process

root 54951 54825 0 18:16 ? 00:00:00 nginx: master process
/opt/local/contrib/sbin/nginx -p /var/log/nginx
www-data 54952 54951 0 18:16 ? 00:00:00 nginx: worker process

www-data 54953 54951 0 18:16 ? 00:00:00 nginx: worker process

Posted at Nginx Forum:

Hello!

On Wed, Nov 03, 2010 at 03:22:53AM -0400, megamic wrote:

Shouldent the original process die after it starts the new process? I
see another server start up, but the old one hang around:

No. It should hang around. Moreever, it should still process
requests. If you see new binary successfully started up - you are
free to instruct old process to shutdown (actually, you are free
to do it even if you don’t see new binary started up - but you
probably don’t want to).

Maxim D.