And is it really necessary to install Git just to download some software? The devs should make a tarball out of the 0.3.0 tag and follow standard packaging conventions. /opt/ is for 3rd-party/proprietary stuff or "really big crap" like KDE/Gnome (if you're on certain distros).
If the generation gap between modern developers and Linux package-managing folk is too big, i'll gladly write up some guides that explain how (and why) to package software instead of slapdash local installs.
For what it's worth, I try to maintain RPM packages for graphite for RHEL (and derivatives) over at http://pakk.96b.it/
If it's supposed to make packages for random unpacked software i'm kind of lost. It could be I just really suck at reading Ruby (don't know the language) but I couldn't find any code that looks at autoconf/automake files or standard open-source developer conventions to generate a package from scratch using the values intended by the developer. I've written two very crappy tools that do this to automatically generate packages (mainly for Solaris, RedHat and Slackware).
To answer the guy above's question: you need to learn how to make basic Makefiles and after that it's all package-manager-specific stuff, which you can use something like Alien (or fpm) to convert to whatever format you wish.
Also, see "Use Case - Package something that uses 'make install'" on the fpm wiki at https://github.com/jordansissel/fpm/wiki/PackageMakeInstall
http://psydev.syw4e.info/new/autopkg.pl/autopkg.pl-1.5/usr/b... (oh god, this is bad code)
You're right about the guide being a little too long. I was trying to establish a starting point for people looking to install and set up these tools but it would be better if there was a way to just "apt-get install" this stuff.
I'll see if I can put this together in a deb and release it somehow.
In short a measurement is a single datapoint i.e. <key, value, timestamp>. So if you have a sense of how many metrics you want to track and at what frequency, it's relatively straight-forward to calculate the list price. There's an estimator on the pricing page to help you do that once you understand what constitutes a "measurement".
That said, librato seems like it would be nice if your stat volume was low.
pattern = ^stats_counts\..*
xFilesFactor = 0.25
aggregationMethod = sum
I sticked to Munin for the time being.
It's very cool that all my metrics are in one place. Shame it is such an ugly, uncomfortable place.