A backend library to be used with a Cube (github.com/square/cube) Collector.
Switch branches/tags
Nothing to show
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.



Major credit to dynmeth's Mongo backend (http://github.com/dynmeth/mongo-statsd-backend).

A backend StatsD library to be used with a Cube (http://github.com/square/cube) Collector. Backend library uses the Cube emitter to send data received when statsd data flushes. Use Cube Evaluator queries (http://github.com/square/cube/wiki/Queries) to pull and use data where needed. Combine with Cubism (http://github.com/square/cubism) for more entertainment. You have to have an operational Cube server, which requires NodeJS and MongoDB to run. Follow Square's installation documents as they're pretty straight forward. Additionally, test your connectivity with the emitter library which would look similar to this

var emitter = require('cube').emitter("ws://");

    type: "type_name",
    time: Date.now(),
        data: "you",
        will: "query"
        from: 150


From your statsd directory

    npm install cube-statsd-backend

Add the backend and a cube emitter web socket DSN to your statsd config file, you probably additionally want to set deleteIdleStats to true so you don't fill cube with a bunch of empty data as the evaluator will take care of filling in the spaces for you already. Since Cube's evaluators will take care of creating metrics it really only stores the stat type and the raw data and throws away any of the flush interval sent across by statsd

    "deleteIdleStats": true,
    "cube": {
        "dsn": "ws://",
        "typePrefix": "prefix_string" //this will be prepended to the mongo collection so care on naming only [A-Za-z_]
    "backends": ['./node_modules/cube-statsd-backend/lib/index.js']


Each collection will be prefixed by the optional typePrefix and with the type of statsd data that was sent (gagues, counters, etc). Make requests to the evaluator api to pull data. StatsD will send along the interval data but the evaluator will take care of a lot of that for you so we just put all data under a single collection instead of appending like


Example Data Objects

To write proper cube expressions to pull your data you will need to know about statsd sends for each counter but here's three examples to help you get started.

Counter: Event with 15 sent for a count in one flush, usually sent with count()/increment()/decrement()

    "count": 15

Timer: Event with five 250ms timers sent in one flush

    "mean_90": 250,
    "upper_90": 250,
    "sum_90": 1250,
    "std": 0,
    "upper": 250,
    "lower": 250,
    "count": 5,
    "count_ps": 5,
    "sum": 1250,
    "mean": 250,
    "median": 250

Gauge: Event with a gauge set to 25 in one flush

    "gauge": 25


0.0.1: fixing for gauges and adding more documentation

0.0.0: all types are operational, but counters are all that I've played with. I'm ignoring bad lines and packets received for now until I know what to do with them.