A Python implementation of Etsy's statsd that supports all the Graphite data types
Latest commit 83851f7 May 22, 2012 @armon armon Adding deprecation warning


NOTE: This version has been deprecated in favor of https://github.com/armon/statsite, which provides API compatibility but is much faster and more scalable.


This is a stats aggregation server. By default, Statsite flushes data to Graphite. Statsite is based heavily on Etsy's StatsD.


  • Basic key/value metrics
  • Send timer data, Statsite will calculate: - Mean - Min/Max - Standard deviation - All the above metrics for a specific percentile of information
  • Send counters that Statsite will aggregate
  • Send a sample rate with counters and Statsite will take that into account when aggregating.


Install Statsite from PyPi:

pip install statsite

Or download and install from source:

python setup.py install

Note: Tested only with Python 2.6+


Statsite preferably should be configured using a file, although all configuration parameters can be set via the command line as well. Here is an example configuration file:

# Settings for the "collector" which is the UDP listener
host =
port = 8125

# Specify settings for the metrics "store" which is where graphite is
host =
port = 2003

Then run statsite, pointing it to that file (assuming /etc right now):

statsite -c /etc/statsite.conf


By default, Statsite will listen for UDP packets, which makes it extremely cheap for your application to fire and forget packets to the server. A message looks like the following (where the flag is optional):


Messages should be separated by newlines (n) if multiple are sent in the same packet.

Currently supported message types:

  • kv - Simple Key/Value. If a flag is given, it is considered the timestamp of the key/value pair.
  • ms - Timer. If a flag is given, it is considered the sampling rate of the timer.
  • c - Counter. After the flush interval, the counters of the same key are aggregated and this is sent to the store.


The following is a simple key/value pair, in this case reporting how many queries we've seen in the last second on MySQL:


The following is a timer, timing the response speed of an API call:


The following is another timer, but this time saying we sample this data in 1/10th of the API requests.


The next example is increments the "rewards" counter by 1:


And this example decrements the "inventory" counter by 7:


As said earlier, multiple messages can be joined together by newlines.