Fetching contributors…
Cannot retrieve contributors at this time
117 lines (75 sloc) 4.18 KB


Simple event based statsd implementation written in python using eventlet with support for reporting to graphite using the pickle protocol.


statsdpy sample config with default options:

#graphite host and port to connect too
graphite_host =
graphite_port = 2003
graphite_pickle_port = 2004

#address and port we should listen for udp packets on
listen_addr =
listen_port = 8125

#If you track a large number of metrics you can use the pickle protocol
pickle_protocol = no
#max number of metrics to report in one go when using the pickle protocol
pickle_batch_size = 300

#enabling debug mode and running in the foreground (with -f) is a great way
#to debug an app generating new metrics
debug = no

#How often to flush stats to graphite
flush_interval = 10
#calculate the XXth percentile
percent_threshold = 90

# Override the key prefixes and suffixes - see
legacy_namespace = true
global_prefix = stats
prefix_counter = counters
prefix_timer = timers
prefix_gauge = gauges
  • Edit the config file appropriately for your environment
  • Start the service: statsdpy-server start --conf=/path/to/your.conf
  • Fire some udp counter, timer, or gauge events at statsdpy
  • Check syslog for any errors starting up or processing events
  • Profit!

Metric Namespacing

Important Note: We default to legacy namespacing at the moment but will probably switch in the near future - see Metric namespacing for further information.

Reporting using the pickle protocol

If you track a decent # of metrics you wanna switch to report to graphite using the pickle protocol. The pickle protocol is more efficient than the the plaintext protocol, and supports sending batches of metrics to carbon in one go. To enable it just set pickle_protocol to "yes" in your statsdpy.conf. Optionally, you can also adjust the max number of items per batch that is reported by adjusting the pickle_batch_size conf option.

Event Types

As with the original statsd implementation from etsy the following event types are supported:



Simple counters. Add 1 to the "sitelogins" event bucket. It stays in memory until flushed to graphite as specified by the flush_interval.


Another counter, this time add "7" to the "500errors" event bucket.



The "pageload" event took 320ms to complete this time. statsdpy computes the XXth percentile (as specified in the config), average (mean), lower and upper bounds for the configured flush interval.

Gauge (simple arbitrary values)####


Combined Events

The etsy statsd implementation now supports combined events via seperation by newline. Statsdpy supports this method now as well:




This counter is being sent sampled every 1/10th of the time.


This counter is being sampled at a 50% rate.


  • eventlet

Building .deb packages

Clone the repo and build the package with stdeb (sudo apt-get install python-stdeb):

git clone
cd statsdpy
python --command-packages=stdeb.command bdist_deb
dpkg -i deb_dist/python-statsdpy_0.0.X-1_all.deb

Installation via

  • git clone
  • cd statsdpy
  • python install
  • Copy the sample config to /etc/statsdpy/statsdpy.conf
  • Edit /etc/statsdpy/statsdpy.conf as required for your environment
  • Start statsdpy /usr/bin/statsdpy-server --conf=/etc/statsdpy/statsdpy.conf start
  • Optionally, a basic init script is provided as etc/statsdpy/statsdpy.init