I am using nginx (free version on centos) as a reverse proxying load balancer to a cluster of Tomcat 7 servers. It works great except for the time when tomcat is starting.
If a Tomcat instance is shut down, it fails over to another correctly. But if Tomcat is starting up - which can take 60 sec - it sends the requests to tomcat but tomcat does not respond. So the user gets the nginx error page.
Is there any way for the upstream load balancer to make sure tomcat is responding?