ControlFreak is a process supervisor. It consists in a set of pure Perl classes, a controlling process usually running in the background and a command line tool to interact with it.
Instances of this main ControlFreak class are called controller,
The supervisor/controller process is running in an EventLoop and forks to start (exec) the services it controls.
Advanced ControlFreak features include:
COW effect: Memory sharing among workers using unix's Copy-On-Write.
It is not a replacement for the init process, init.d etc... The initial goal of ControlFreak is to simplify the management of all the processes required to run a modern web application. An average web app would use:
A web reverse proxy or balancer, like Perlbal
Multiple kind of workers
A web server or an application server (apache, fastcgi, ...)
More complex environments add a lot of additional services.
In production you want to tightly control those, making sure there are up and running nominally. You also want an easy way to do code pushes and soft roll releases.
In development you usually want to duplicate the production stack which is a lot of services that you have to tweak and sometimes restart repeatedly, and be able to slightly tweak based on the developer, the code branch etc...
In test, you want a few of these services, and you want to programatically control them (making sure there are up or down)
Pid management is always a nightmare when you want to cover all these needs.
There are many similar programs freely available, but as stated above, ControlFreak does a few things differently (and hopefully better). ControlFreak is written in Perl which makes it crash resistant and easy to debug/inspect.
ControlFreak is also designed to be as simple as permitted. In order to keep the core
ControlFreak stable, easy to understand and easy to run, there should be little added to the core features of running services and providing means to control them. For instance,
ControlFreak doesn't have on the roadmap to develop features to auto-restart process if they use too much memory, or to email you when some process has a snafu. We believe you can have another script/process interfacing with
ControlFreak which does exactly that.
ControlFreak wants to do one thing and only one thing well, and that thing is: run services, provide means to control their lifecycle.
From the above, it results, that ControlFreak is no:
memory bloat watcher
application data store
service restart scheduler
event bus for your services
The memory overhead is minimal considering it uses perl. On my 64bits linux, here is the weight of
yann@xen:~/git/ControlFreak$ ps -p 27554 -o size= 12948 yann@laptop:~$ ps -p 16452 -o rss RSS 5896