statsite

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

Install

Install Statsite from PyPi:

pip install statsite

Or download and install from source:

python setup.py install

Usage

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
[collector]
host = 0.0.0.0
port = 8125

# Specify settings for the metrics "store" which is where graphite is
[store]
host = 0.0.0.0
port = 2003

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

statsite -c /etc/statsite.conf

Protocol

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):

key:value|type[|@flag]

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.

Examples:

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

mysql.queries:1381|kv|@1313107325

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

api.session_created:114|ms

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

api.session_created:114|ms|@0.1

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

rewards:1|c

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

inventory:-7|c

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

Features:

• 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.

Requirements:

• Python