Find list of processes and connections:
lsof -i
Kill worker processes
sudo pkill nginx
Restart nginx
sudo service nginx restart
Good logs to check:
/var/log/nginx/error.log
/var/log/upstart/consul-template.log
(side note for me, cause consul<-->nginx closely related)
@leohowell, there are no email notifications for gist comments so sorry I didn't see this earlier. In my case the cause was that I was applying many rapid change to my NGINX configuration and calling:
This eventually put the master process in a weird state and it died, this leave the child processes as zombies. This behavior is documented in an unresolved NGINX issue: yaoweibin/nginx_upstream_check_module#53.