A utility for creating gauges from prometheus counters and histograms in the python client.
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
prometheus_roller
tests
.gitignore
.travis.yml
LICENSE
README.md
setup.cfg
setup.py

README.md

Prometheus Roller

Build Status PyPI version Downloads Python versions License

This is a small helper utility for creating time-boxed rollups for histogram and counter metrics created using the prometheus python client.

Metrics like counters and histograms increase indefinitely, but you usually care about recent activity. This library provides a way to turn a counter or histogram into a gauge or set of gauges that give instantaneous measures of the state of the application.

This isn't very useful if you're using the prometheus server since it provides functions to do that for you, but if you are sending metrics to multiple places (e.g. check_mk) and you want to send those metrics in a form that make basic alerting and reporting easier, this may help.

Usage

from prometheus_client import Histogram, Counter
from prometheus_roller import HistogramRoller, CounterRoller, start_update_daemon

# Create a histogram
h = Histogram('test_value', 'Testing roller')

# Create a counter
c = Counter('test_counted_value', 'Testing roller')

# Create a roller for the histogram, which calculates windowed values.
# By default it will create a gauge with a label for each histogram bin.
# The value of each gauge will be the change in value over the last 5 minutes, updated every 5 seconds.
# See the `options` parameter for more configuration options.
rh = HistogramRoller(h)

# Create a roller for the counter, which calculates windowed values.
# The value of each gauge will be the change in value over the last 5 minutes, updated every 5 seconds.
# See the `options` parameter for more configuration options.
rc = CounterRoller(c)

# Launch a daemon thread tracking and updating all roller objects.
# See the code for more options for configuring this update process.
start_update_daemon()

Installation

# Install with pip
pip install prometheus_roller

# To install as editable to work on it
pip install -e git+https://github.com/turtlemonvh/prometheus_python_roller.git#egg=prometheus_python_roller
# OR
git clone git@github.com:turtlemonvh/prometheus_python_roller.git prometheus-roller
cd prometheus-roller
python setup.py develop

Running tests

# Plain old python
python -m unittest discover

# If you have nose and coverage installed
nosetests --with-cover

TODO

  • Add IQR reducer