In my latest frenzy, which was focused on HA more than performance, I installed some new servers, new services on those servers, and the general complexity of the entire setup for beeets.com doubled. I was trying to remember a utility that I saw a while back that would restart services if they failed. I checked my delicious account, praying that I had thought of my future self when I originally saw it. Luckily, I had saved it under my "linux" tag. Thanks, Andrew from the past.

The tool is called monit, and I'm surprised I ever lived without it. Not only does it monitor your services and keep them running, it can restart them if they fail, use too much memory/cpu, stop responding on a certain port, etc. Not only that, but it will email you every time something happens.

While perusing monit's site, I saw M/Monit which allows you to monitor monit over web, essentially. The only thing I scratched my head about was that M/Monit uses port 8080 (which is fine) but NginX already uses port 8080, and I wasn't about to change that, so I opened conf/server.xml and looked for 8080, replaced with 8082 (monit runs on 8081 =)). Then I reconfigured monit to communicate with M/Monit and vice versa, and now I have a kickass process monitor that alerts me when things go wrong, and also sends updates to a service that allows me to monitor the monitor.

I can't look at things like queries/sec as I can with Cacti (which is awesome but a little clunky) but I can see which important services are running on each of my servers, and even restart them if I need to straight from M/Monit. The free download license allows to use M/Monit on one server, which is all I need anyway.

Great job monit team, you have gone above and beyond.