Try using "reload". There should be no downtime with that.
If you are using the latest version (0.7.53) you can simply run:
# nginx -s reload
If you are on an earlier version, and are running as a Linux service then you can probably use
# service nginx reload
You can confirm that the master process has not been terminated but the workers all have new PID's with
# ps -e | grep nginx
--
Jim Ohlstein