PyMetrics

Redis backed metrics library - implements the most of the famous Metrics library.

Classes

RetricsGauge(BaseMetrics) - Single value gauge
RetricsCounter(BaseMetrics) - Simple counter with incr and decr methods
RetricsMeter(BaseMetrics) - Time series data, with 1, 5 and 15 minutes avg
RetricsHistogram(BaseMetrics) - Histogram with percentile, mean, median and std deviation methods 
RetricsTimer(BaseMetrics) - Timer (wallclock)

The main class to loof after is RetricsFactory - Metrics factory 

Hierarchy

Basically we register an application and its metrics instances in the following order:
    Application -> Metrics -> Instances of metrics

The important thing to monitor is that each metric will have an internal name based on the application + metric name + pid.
By looking at the way the name is composed it's easy to interchange data between processes.

Examples:

from pymetrics import RetricsFactory

rf = RetricsFactory('application_name')                                                       
c = rf.new_counter('requests')                                              
c.incr()                                                                    
c.decr()

To list all instances for a given metric:

rf.list_instances_per_metric('gauge') 

rf.unregister_instance(c)

Unregister metrics instances is not mandatory as each new metric will not use the same internal name.

For more examples, check the tests/ directory    
Use python -m unittest discover tests/unit to run all tests

TODO

Should use riemann optionally
Should integrate with dashify 
Should be ported to ruby