"Peter Brezny" <[EMAIL PROTECTED]> writes:
> What are the primary advantages of using supervise?
Among those already mentioned: reliability. You *can't* reliably
manage a service without cooperation from the parent process or the
process itself. Putting the management functionality into the process
itself results in unnecessary duplication; putting it in the parent
results in supervise.
An example of the kind of unreliability you get without supervise:
suppose you want to send a service a signal. How do you find the pid?
ps? Command names are not a perfect indicator; multiple instances
complicate the problem. A pid file? It might be out of date: the
process might have died since the pid file was written, and the pid
might have been reused. But a parent can always keep track of its
children; supervise never sends signals to the wrong process.
paul
