Skip to main content
Python Software Foundation 20th Year Anniversary Fundraiser  Donate today!

A library to enable collection and delayed emission of StatsD metrics using the DataDog protocol.

Project description

dogstatsd-collector is a library to make it easy to collect DataDog-style StatsD counters and histograms with tags and control when they are flushed. It gives you a drop-in wrapper for the DogStatsD library for counters and histograms and allows you to defer flushing the metrics until you choose to. This capability enables you to collect StatsD metrics at arbitrary granularity, for example on a per-web request or per-job basis (instead of the per-flush interval basis).

Counters and histograms are tracked separately for each metric series (unique set of tag key-value pairs) and a single metric is emitted for each series when the collector is flushed. You don’t have to think about tracking your metric series separately; you just use the DogstatsdCollector object as you would the normal DogStatsD object, and flush when you’re ready; the library will take care of emitting all the series for you.

  • Free software: BSD 3-Clause License


pip install dogstatsd-collector

Example Usage

Imagine you want to track a distribution of the number of queries issued by requests to your webapp, and tag them by which database is queried and which verb is used. You collect the following metrics as you issue your queries:

collector = DogstatsdCollector(dogstatsd)
collector.histogram('query', tags=['database:master','verb:insert'])
collector.histogram('query', tags=['database:master','verb:update'])
collector.histogram('query', tags=['database:master','verb:update'])
collector.histogram('query', tags=['database:replica','verb:select'])
collector.histogram('query', tags=['database:replica','verb:select'])

Then, at the end of your web request, when you flush the collector, the following metrics will be pushed to DogStatsD (shown in DogStatsD datagram format):

# query:1|h|#database:master,verb:insert
# query:2|h|#database:master,verb:update
# query:2|h|#database:replica,verb:select

Base Tags

The collector object also supports specifying a set of base tags, which will be included on every metric that gets emitted.

base_tags = ['mytag:myvalue']
collector = DogstatsdCollector(dogstatsd, base_tags=base_tags)
collector.histogram('query', tags=['database:master','verb:insert'])
collector.histogram('query', tags=['database:master','verb:update'])
# query:1|h|#database:master,verb:insert,mytag:myvalue
# query:1|h|#database:master,verb:update,mytag:myvalue


The StatsD model is to run an agent on each server/container in your infrastructure and periodically flush aggregations at a regular interval to a centralized location. This model scales very well because the volume of metrics sent to the centralized location grows very slowly even as you scale your application; each StatsD agent calculates aggregations to flush to the backend instead of every datapoint, so the storage volume is quite low even for a large application with lots of volume.

A drawback to this model is that you don’t have much control of the granularity that your metrics represent. When your aggregations reach the centralized location (DataDog in this case), you only know the counts or distributions within the flush interval. You can’t represent any other execution granularity beyond “across X seconds” (where X is the flush interval). This limitation precludes you from easily representings metrics on a “per-request” basis, for example.

The purpose of this library is to make it simple to control when your StatsD metrics are emitted so that you can defer emission of the metrics until a point you determine. This allows you to represent a finer granularity than “across X seconds” such as “across a web request” or “across a cron job.” It also preserves metric tags by emitting each series independently when the collector is flushed, which ensures you don’t lose any of the benefit of tagging your metrics (such as aggregating/slicing in DataDog).


The DogstatsdCollector object is a singleton that provides a similar interface as the DogStatsD increment and histogram methods. As you invoke these methods, you collect counters and histograms for each series (determined by any tags you include). After calling flush(), each series is separately emitted as a StatsD metric.

Simple Request Metrics

You can collect various metrics over a request and emit them at the end of the request to get per-request granularity.

In Django:

from datadog.dogstatsd.base import DogStatsd
from dogstatsd_collector import DogstatsdCollector

# Middleware
class MetricsMiddleware:
    def __init__(self, get_response):
        self.get_response = get_response
        self.dogstatsd = DogStatsd()

    def __call__(self, request):
        request.metrics = DogstatsdCollector(self.dogstatsd)
        response = self.get_response(request)

        return response

# Inside a view
def my_view(request):
    # Do some stuff...
    request.metrics.histogram('my.time', 0.5)
    return HttpResponse('ok')

In Flask:

from datadog.dogstatsd.base import DogStatsd
from dogstatsd_collector import DogstatsdCollector

from flask import Flask
from flask import request

app = Flask(__name__)
dogstatsd = DogStatsd()

def init_metrics():
    request.metrics = DogstatsdCollector(dogstatsd)

def flush_metrics():

def my_view():
    # Do some stuff...
    request.metrics.histogram('my.time', 0.5)
    return 'ok'

Celery Task Metrics

Same as above, but over a Celery task.

from datadog.dogstatsd.base import DogStatsd
from dogstatsd_collector import DogstatsdCollector

from celery import Celery
from celery import current_task
from celery.signals import task_prerun
from celery.signals import task_postrun

app = Celery('tasks', broker='pyamqp://guest@localhost//')

dogstatsd = DogStatsd()

def init_metrics(task_id, task, *args, **kwargs):
    task.request.metrics = DogstatsdCollector(dogstatsd)

def flush_metrics(task_id, task, *args, **kwargs):

def my_task():
    # Do some stuff...
    current_task.request.metrics.histogram('my.time', 0.5)

Metrics Within a Function

Emit a set of metrics for a particular function you execute.

from datadog.dogstatsd.base import DogStatsd
from dogstatsd_collector import DogstatsdCollector

dogstatsd = DogStatsd()

def do_stuff(metrics):
    # Do some stuff...
    metrics.histogram('my.time', 0.5)

metrics = DogstatsdCollector(dogstatsd)

Thread Safety

The DogstatsdCollector singleton is not threadsafe. Do not share a single DogstatsdCollector object among multiple threads.

More Documentation

Full documentation can be found on ReadTheDocs:


To run the all tests run:



0.0.2 (2019-08-14)

  • Add base_tags optional kwarg to support tags added to all metrics that get flushed.

0.0.1 (2019-05-02)

  • First release on PyPI.

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for dogstatsd-collector, version 0.1.0
Filename, size File type Python version Upload date Hashes
Filename, size dogstatsd-collector-0.1.0.tar.gz (17.8 kB) File type Source Python version None Upload date Hashes View

Supported by

AWS AWS Cloud computing Datadog Datadog Monitoring DigiCert DigiCert EV certificate Facebook / Instagram Facebook / Instagram PSF Sponsor Fastly Fastly CDN Google Google Object Storage and Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Salesforce Salesforce PSF Sponsor Sentry Sentry Error logging StatusPage StatusPage Status page