A DogStatsd Python client.

First install the library with pip or easy_install

# Install in system python ...
sudo easy_install dogstatsd-python

# .. or into a virtual env
easy_install  dogstatsd-python

Then start instrumenting your code:

# Import the module.
from statsd import statsd

# Optionally, configure the host and port if you're running Statsd on a
# non-standard port.
statsd.connect('localhost', 8125)

# Increment a counter.
statsd.increment('page.views')

# Record a gauge 50% of the time.
statsd.gauge('users.online', 123, sample_rate=0.5)

# Sample a histogram.
statsd.histogram('file.upload.size', 1234)

# Time a function call.
@statsd.timed('page.render')
def render_page():
    # Render things ...

# Tag a metric.
statsd.histogram('query.time', 10, tags = ["version:1"])

You can also post events to your stream. You can tag them, set priority and even aggregate them with other events.

Aggregation in the stream is made on hostname/event_type/source_type/aggregation_key.

# Post a simple message
statsd.event("There might be a storm tomorrow", "A friend warned me earlier.")

# Cry for help
statsd.event("SO MUCH SNOW", "Started yesterday and it won't stop !!", alert_type = "error", tags = ["urgent", "endoftheworld"])

Read the full API docs here.

To suggest a feature, report a bug, or general discussion, head over here.

• 0.5.3
  • Add support for service checks
• 0.5.2
  • Recreate a socket on SocketError to not stay stuck
• 0.5.1
  • Add support for batch sending metrics
• 0.5.0
  • Encode Binary payload with UTF-8
• 0.4.1
  • Re-add global statsd instance (thanks to @Julian).
• 0.4.0
  • Added support for sending events.
• 0.3.0
  • Uses a connected socket for a big performance improvement (thanks to @Julian).
  • Use connect to override the host and port of a statsd instance.
• 0.2.1
  • Fixed the timed decorator, to ensure it preserves function attributes.
• 0.2
  • Added the set metric type.
• 0.1
  • Initial version of the code

Copyright (c) 2015 Datadog Inc.