microsoft
/
statsd
зеркало из https://github.com/microsoft/statsd.git
1
0
Форкнуть 0
Код Задачи Пакеты Проекты Релизы Вики Активность
Simple daemon for easy stats aggregation
25 коммитов 6 Ветки 11 Теги 784 KiB
JavaScript 97.1%
Shell 1.8%
Perl 1.1%
Перейти к файлу
Steve Ivy b40d55c7a7 fix example code for timing. 2011-04-12 02:21:08 +08:00
debian Perms change 2011-04-12 02:18:44 +08:00
LICENSE adding license (mit) 2010-12-29 19:50:11 -05:00
README.md Typo in README.md 2011-04-12 02:17:16 +08:00
StatsdClient.java Make socket reusable to avoid "Too many open files". Clean up constructors. 2011-04-12 02:19:58 +08:00
config.js initial import of StatsD 2010-12-29 19:15:00 -05:00
exampleConfig.js initial import of StatsD 2010-12-29 19:15:00 -05:00
php-example.php initial import of StatsD 2010-12-29 19:15:00 -05:00
python_example.py fix example code for timing. 2011-04-12 02:21:08 +08:00
stats.js don\'t die when receiving unusual data 2011-04-12 02:16:35 +08:00

README.md

StatsD

A network daemon for aggregating statistics (counters and timers), rolling them up, then sending them to graphite.

We (Etsy) blogged about how it works and why we created it.

Concepts

  • buckets Each stat is in it's own "bucket". They are not predefined anywhere. Buckets can be named anything that will translate to Graphite (periods make folders, etc)

  • values Each stat will have a value. How it is interpreted depends on modifiers

  • flush After the flush interval timeout (default 10 seconds), stats are munged and sent over to Graphite.

Counting

gorets:1|c

This is a simple counter. Add 1 to the "gorets" bucket. It stays in memory until the flush interval.

Timing

glork:320|ms

The glork took 320ms to complete this time. StatsD figures out 90th percentile, average (mean), lower and upper bounds for the flush interval.

Sampling

gorets:1|c|@0.1

Tells StatsD that this counter is being sent sampled every 1/10th of the time.

Guts

  • UDP Client libraries use UDP to send information to the StatsD daemon.

  • NodeJS

  • Graphite

Graphite uses "schemas" to define the different round robin datasets it houses (analogous to RRAs in rrdtool). Here's what Etsy is using for the stats databases:

[stats]
priority = 110 
pattern = ^stats\..*
retentions = 10:2160,60:10080,600:262974

That translates to:

  • 6 hours of 10 second data (what we consider "near-realtime")
  • 1 week of 1 minute data
  • 5 years of 10 minute data

This has been a good tradeoff so far between size-of-file (round robin databases are fixed size) and data we care about. Each "stats" database is about 3.2 megs with these retentions.

Inspiration

StatsD was inspired (heavily) by the project (of the same name) at Flickr. Here's a post where Cal Henderson described it in depth: Counting and timing. Cal re-released the code recently: Perl StatsD

Contribute

You're interested in contributing to StatsD? AWESOME. Here are the basic steps:

fork StatsD from here: http://github.com/etsy/statsd

  1. Clone your fork
  2. Hack away
  3. If you are adding new functionality, document it in the README
  4. If necessary, rebase your commits into logical chunks, without errors
  5. Push the branch up to GitHub
  6. Send a pull request to the etsy/statsd project.

We'll do our best to get your changes in!

Contributors

In lieu of a list of contributors, check out the commit history for the project: http://github.com/etsy/statsd/commits/master