Awesome

TxMetrics

Redis backed metrics library ported to twisted. Compatible with https://github.com/gleicon/pymetrics. Implements part of the famous Metrics library. Data is interchangeable between both libraries (you can have a traditional python app with pymetrics and a twisted based app sharing data, etc)

Classes

TxMetricsGauge(BaseMetrics) - Single value gauge
TxMetricsCounter(BaseMetrics) - Simple counter with incr and decr methods
TxMetricsMeter(BaseMetrics) - Time series data, with 1, 5 and 15 minutes avg
TxMetricsHistogram(BaseMetrics) - Histogram with percentile, mean, median and std deviation methods 
TxMetricsTimer(BaseMetrics) - Timer (wallclock)

The main class to look for is TxMetricsFactory - 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:

There is an example/ directory with the general outline for each kind of metric.
Run all tests with trial tests/
All examples used redis.ConnectionPool() as connector.
If your application happens to use lazyConnection or lazyConnectionPool, you need to wait for the conn to be true.
To do that, add the following code in your method (surrounded by defer.inlineCallbacks):
    r = redis.lazyConnectionPool(....)
    yield r._connected

Depends on

Twisted and txredisapi https://github.com/fiorix/txredisapi or the cyclone redis driver