What Can Graphite Tell You About Your Servers?

Graphite is a time series data storage and graphing system.

OK, what does that mean?

That means you can shove numbers at it over time, and it will store them very efficiently and draw graphs with them.

For example, let's pick a particular metric; say you want to monitor the load on a server. All you have to do is send that number, with an identifying string and a timestamp, to Graphite and it will store it. Obviously, time series data can grow forever, so Graphite is very configurable as to how long it will keep certain resolutions of data, and how to "roll them up " for lower resolution storage. For example, storing a single metric with 10 second resolution for one year and 1 minute resolution for 10 years requires [only] 13MB of space.

[Re]Consider Your "Default" Configuration(s)

A colleague (I'll call him Chris) called me a couple of days ago asking if I had a few minutes. He said he had a client who was speaking at a conference and promoting his software; he ended up on Hacker News, and his site was now somewhere between dying and dead. To add insult to injury, he was annoucing that his company was open sourcing their database monitoring software, and their website was displaying "Unable to establish database connection".

Ouch!