Measured Reporting

The registry and reporting library that has the classes needed to create a dimension aware, self reporting metrics registry.

Install

npm install measured-reporting

What is in this package

Self Reporting Metrics Registry

A dimensional aware self-reporting metrics registry, just supply this class with a reporter implementation at instantiation and this is all you need to instrument application level metrics in your app.

See the SelfReportingMetricsRegistryOptions for advanced configuration.

const { SelfReportingMetricsRegistry, LoggingReporter } = require('measured-reporting');
const registry = new SelfReportingMetricsRegistry(new LoggingReporter({
  defaultDimensions: {
    hostname: os.hostname()
  }
}));
 
// The metric will flow through LoggingReporter#_reportMetrics(metrics) every 10 seconds by default
const myCounter = registry.getOrCreateCounter('my-counter');
 

Reporter Abstract Class

Extend this class and override the _reportMetrics(metrics) method to create a vendor specific reporter implementation.

See the ReporterOptions for advanced configuration.

Current Implementations

• SignalFx Reporter in the measured-signalfx-reporter package.
  • reports metrics to SignalFx.
• Logging Reporter in the measured-reporting package.
  • A reporter impl that simply logs the metrics via the Logger

Creating an anonymous Implementation

You can technically create an anonymous instance of this, see the following example.

const os = require('os');
const process = require('process');
const { SelfReportingMetricsRegistry, Reporter } = require('measured-reporting');
 
// Create a self reporting registry with an anonymous Reporter instance;
const registry = new SelfReportingMetricsRegistry(
  new class extends Reporter {
    constructor() {
      super({
        defaultDimensions: {
          hostname: os.hostname(),
          env: process.env['NODE_ENV'] ? process.env['NODE_ENV'] : 'unset'
        }
      })
    }
 
    _reportMetrics(metrics) {
      metrics.forEach(metric => {
        console.log(JSON.stringify({
          metricName: metric.name,
          dimensions: this._getDimensions(metric),
          data: metric.metricImpl.toJSON()
        }))
      });
    }
  }()
);
 
// create a gauge that reports the process uptime every second
const processUptimeGauge = registry.getOrCreateGauge('node.process.uptime', () => process.uptime(), {}, 1);

Example output:

APP5HTD6ACCD8C:foo jfiel2$ NODE_ENV=development node index.js
{"metricName":"node.process.uptime","dimensions":{"hostname":"APP5HTD6ACCD8C","env":"development"},"data":0.092}
{"metricName":"node.process.uptime","dimensions":{"hostname":"APP5HTD6ACCD8C","env":"development"},"data":1.099}
{"metricName":"node.process.uptime","dimensions":{"hostname":"APP5HTD6ACCD8C","env":"development"},"data":2.104}
{"metricName":"node.process.uptime","dimensions":{"hostname":"APP5HTD6ACCD8C","env":"development"},"data":3.105}
{"metricName":"node.process.uptime","dimensions":{"hostname":"APP5HTD6ACCD8C","env":"development"},"data":4.106}

Consider creating a proper class and contributing it back to Measured if it is generic and sharable.

Logging Reporter Class

A simple reporter that logs the metrics via the Logger.

See the ReporterOptions for advanced configuration.

const { SelfReportingMetricsRegistry, LoggingReporter } = require('measured-reporting');
const registry = new SelfReportingMetricsRegistry(new LoggingReporter({
  logger: myLogerImpl, // defaults to new console logger if not supplied
  defaultDimensions: {
    hostname: require('os').hostname()
  }
}));

What are dimensions?

As described by Signal Fx:

A dimension is a key/value pair that, along with the metric name, is part of the identity of a time series. You can filter and aggregate time series by those dimensions across SignalFx.

DataDog has a nice blog post about how they are used in their aggregator api.

Graphite also supports the concept via tags.