I don't see a monitoring system there? Something like Supervisor is great, but who watches the watcher? I prefer Nagios, but that's just because I'm used to it. Even it is better than nothing.

If you want something a bit more distributed (read less prone to single points of failure) than Supervisor, check out pacemaker. Very powerful and useful for keeping resources alive in any sized cluster.