Working on building a High Availability system for a customer.

The goal is to provide fail-over services for our system which includes our own proprietary application that runs on top of httpd/postgresql/samba/tomcat. These systems run on CentOS 4 servers.

I have set up the drbd side so far and it syncs fine between the two systems. I can modify the data, mount, unmount, remount, modify data, etc.

My question is with implementing heartbeat and controlling the services and restoration of the primary system after a failure. I have been reading a ton of documentation online, but I was really looking for someone who has implemented this kind of system(most notably with postgres/httpd) and what potential pitfalls to look out for, what things may be required but are not clearly stated in the documentation?

Does heartbeat actually handle the startup and shutdown of the services like postgres/httpd/samba or should they be configured to startup as normal in the init system? This seems illogical since the drbd partition where the database is located is not mounted the the secondary node...

Any insight is appreciated.

Cheers.