Django App Metrics

django-app-metrics allows you to capture and report on various events in your applications. You simply define various named metrics and record when they happen. These might be certain events that may be immediatey useful, for example 'New User Signups', 'Downloads', etc.

Or they might not prove useful until some point in the future. But if you begin recording them now you'll have great data later on if you do need it.

For example 'Total Items Sold' isn't an exciting number when you're just launching when you only care about revenue, but being able to do a contest for the 1 millionth sold item in the future you'll be glad you were tracking it.

You then group these individual metrics into a MetricSet, where you define how often you want an email report being sent, and to which User(s) it should be sent.

Documentation

Documentation can be found at ReadTheDocs.

Requirements

Celery and django-celery must be installed, however if you do not wish to actually use Celery you can simply set CELERY_ALWAYS_EAGER = True in your settings and it will behave as if Celery was not configured.

Django 1.2 and above

Usage

from app_metrics.utils import create_metric, metric, timing, Timer, gauge

# Create a new metric to track
my_metric = create_metric(name='New User Metric', slug='new_user_signup')

# Create a MetricSet which ties a metric to an email schedule and sets
# who should receive it
my_metric_set = create_metric_set(name='My Set',
                                  metrics=[my_metric],
                                  email_recipients=[user1, user2])

# Increment the metric by one
metric('new_user_signup')

# Increment the metric by some other number
metric('new_user_signup', 4)

# Aggregate metric items into daily, weekly, monthly, and yearly totals
# It's fairly smart about it, so you're safe to run this as often as you
# like
manage.py metrics_aggregate

# Send email reports to users
manage.py metrics_send_mail

# Create a timer (only supported in statsd backend currently)
with timing('mytimer'):
  for x in some_long_list:
     call_time_consuming_function(x)

# Or if a context manager doesn't work for you you can use a Timer class
t = Timer()
t.start()
something_that_takes_forever()
t.stop()
t.store('mytimer')

# Gauges are current status type dials (think fuel gauge in a car)
# These simply store and retrieve a value
gauge('current_fuel', '30')
guage('load_load', '3.14')

Backends

app_metrics.backends.db (Default) - This backend stores all metrics and aggregations in your database. NOTE: Every call to metric() generates a database write, which may decrease your overall performance is you go nuts with them or have a heavily traffic site.

app_metrics.backends.mixpanel - This backend allows you to pipe all of your calls to metric() to Mixpanel. See the Mixpanel documentation for more information on their API.

app_metrics.backends.statsd - This backend allows you to pipe all of your calls to metric() to a statsd server. See statsd for more information on their API.

app_metrics.backends.redis - This backend allows you to use the metric() and gauge() aspects, but not timer aspects of app_metrics.

app_metrics.backends.librato_backend - This backend lets you send metrics to Librato. See the Librato documentation for more information on their API. This requires the Librato library. It uses use a librato Gauge by default, although this can be overridden by supplying metric_type="counter" as a keyword arg to metric().

app_metrics.backends.composite - This backend lets you compose multiple backends to which metric-calls are handed. The backends to which the call is sent can be configured with the APP_METRICS_COMPOSITE_BACKENDS setting. This can be overridden in each call by supplying a backends keyword argument:

metric('signups', 42, backends=['app_metrics.backends.librato',
                                'app_metrics.backends.db'])

Settings

APP_METRICS_BACKEND - Defaults to 'app_metrics.backends.db' if not defined.

APP_METRICS_SEND_ZERO_ACTIVITY - Prevent e-mails being sent when there's been no activity today (i.e. during testing). Defaults to True.

APP_METRICS_DISABLED - If True, do not track metrics, useful for debugging. Defaults to False.

Mixpanel Settings

Set APP_METRICS_BACKEND == 'app_metrics.backends.mixpanel'.

APP_METRICS_MIXPANEL_TOKEN - Your Mixpanel.com API token

APP_METRICS_MIXPANEL_URL - Allow overriding of the API URL end point

Statsd Settings

Set APP_METRICS_BACKEND == 'app_metrics.backends.statsd'.

APP_METRICS_STATSD_HOST - Hostname of statsd server, defaults to 'localhost'

APP_METRICS_STATSD_PORT - statsd port, defaults to '8125'

APP_METRICS_STATSD_SAMPLE_RATE - statsd sample rate, defaults to 1

Redis Settings

Set APP_METRICS_BACKEND == 'app_metrics.backends.redis'.

APP_METRICS_REDIS_HOST - Hostname of redis server, defaults to 'localhost'

APP_METRICS_REDIS_PORT - redis port, defaults to '6379'

APP_METRICS_REDIS_DB - redis database number to use, defaults to 0

Librato Settings

Set APP_METRICS_BACKEND == 'app_metrics.backends.librato'.

APP_METRICS_LIBRATO_USER - Librato username

APP_METRICS_LIBRATO_TOKEN - Librato API token

APP_METRICS_LIBRATO_SOURCE - Librato data source (e.g. 'staging', 'dev'...)

Composite Backend Settings

Set APP_METRICS_BACKEND == 'app_metrics.backends.composite'.

APP_METRICS_COMPOSITE_BACKENDS - List of backends that are used by default, e.g.:

APP_METRICS_COMPOSITE_BACKENDS = ('librato', 'db', 'my_custom_backend',)

Running the tests

To run the tests you'll need some requirements installed, so run:

pip install -r requirements/test.txt

Then simply run:

django-admin.py test --settings=app_metrics.tests.settings

TODO

• Improve text and HTML templates to display trending data well