OpenTelemetry Metrics SDK

This module contains the Metrics SDK of opentelemetry-js.

Used standalone, this module provides methods for manual instrumentation of code, offering full control over recording metrics for client-side JavaScript (browser) and Node.js.

It does not provide automated instrumentation of known libraries or host environment metrics out-of-the-box.

Installation

npm install --save @opentelemetry/api
npm install --save @opentelemetry/sdk-metrics
Copy

Usage

The basic setup of the SDK can be seen as followings:

const opentelemetry = require('@opentelemetry/api');
const { MeterProvider } = require('@opentelemetry/sdk-metrics');

// To create an instrument, you first need to initialize the Meter provider.
// NOTE: The default OpenTelemetry meter provider does not record any metric instruments.
//       Registering a working meter provider allows the API methods to record instruments.
opentelemetry.metrics.setGlobalMeterProvider(new MeterProvider());

// To record a metric event, we used the global singleton meter to create an instrument.
const counter = opentelemetry.metrics.getMeter('default').createCounter('foo');

// record a metric event.
// NOTE: By default, each instrument can track up to 2000 unique time series.
//       This can be configured using cardinalityLimits. See "Configuring Cardinality Limits" below.
counter.add(1, { attributeKey: 'attribute-value' });
Copy

In conditions, we may need to setup an async instrument to observe costly events:

// Creating an async instrument, similar to synchronous instruments
const observableCounter = opentelemetry.metrics
  .getMeter('default')
  .createObservableCounter('observable-counter');

// Register a single-instrument callback to the async instrument.
observableCounter.addCallback(async observableResult => {
  // ... do async stuff
  observableResult.observe(1, { attributeKey: 'attribute-value' });
});

// Register a multi-instrument callback and associate it with a set of async instruments.
opentelemetry.metrics
  .getMeter('default')
  .addBatchObservableCallback(batchObservableCallback, [observableCounter]);
async function batchObservableCallback(batchObservableResult) {
  // ... do async stuff
  batchObservableResult.observe(observableCounter, 1, {
    attributeKey: 'attribute-value',
  });
}
Copy

Views can be registered when instantiating a MeterProvider:

const meterProvider = new MeterProvider({
  views: [
    // override the bucket boundaries on `my.histogram` to [0, 50, 100]
    {
      aggregation: {
        type: AggregationType.EXPLICIT_BUCKET_HISTOGRAM,
        options: {
          boundaries: [0, 50, 100],
        },
      },
      instrumentName: 'my.histogram',
    },
    // rename 'my.counter' to 'my.renamed.counter'
    { name: 'my.renamed.counter', instrumentName: 'my.counter' },
  ],
});
Copy

Configuring Cardinality Limits

The cardinalityLimits is an optional property in PeriodicExportingMetricReader that allows configuration of the maximum cardinality limits per instrument type (InstrumentType). This limit controls the maximum number of unique time series that can be tracked for each metric instrument. If not specified in the property, the limit will default to 2000 (the default value can also be specified).

It is converted to a cardinalitySelector function that:

• Takes an InstrumentType as input
• Returns the configured cardinality limit for that instrument type
• Falls back to the default value if a specific type isn't configured
• Uses 2000 as the default if the default value is also not specified

If the cardinalityLimits property is omitted:

import { PeriodicExportingMetricReader } from '@opentelemetry/sdk-metrics';
import { OTLPMetricExporter } from '@opentelemetry/exporter-metrics-otlp-http';

const exporter = new OTLPMetricExporter();
const reader = new PeriodicExportingMetricReader({
  exporter,
  exportIntervalMillis: 60000,
});

// All instruments will use the default limit of 2000 time series
Copy

Configuring specific instrument types:

const reader = new PeriodicExportingMetricReader({
  exporter,
  exportIntervalMillis: 60000,
  cardinalityLimits: {
    counter: 10000, // Counters can have up to 10,000 time series
    histogram: 5000, // Histograms limited to 5,000 time series
    gauge: 3000, // Gauges limited to 3,000 time series
    default: 2500, // changes the default from 2000 to 2500
  },
});
Copy

Available configuration options:

type cardinalityLimits = {
  counter?: number;
  gauge?: number;
  histogram?: number;
  upDownCounter?: number;
  observableCounter?: number;
  observableGauge?: number;
  observableUpDownCounter?: number;
  default?: number;
};
Copy

Configuring Max Export Batch Size

The maxExportBatchSize option in PeriodicExportingMetricReader configures the maximum number of metric data points in a single batch provided to exporter.export(). If omitted, no limit is applied and all collected metric data points are exported in a single batch.

When maxExportBatchSize is configured to a positive integer, PeriodicExportingMetricReader splits the collected metric data into smaller batches of at most maxExportBatchSize data points and exports them sequentially. The exportTimeoutMillis timeout applies to each individual batch export.

import { PeriodicExportingMetricReader } from '@opentelemetry/sdk-metrics';
import { OTLPMetricExporter } from '@opentelemetry/exporter-metrics-otlp-http';

const exporter = new OTLPMetricExporter();
const reader = new PeriodicExportingMetricReader({
  exporter,
  exportIntervalMillis: 60000,
  exportTimeoutMillis: 30000,
  maxExportBatchSize: 200, // Split exports into batches of at most 200 data points
});
Copy

Example

See examples/prometheus for an end-to-end example, including exporting metrics.

Useful links

• For more information on OpenTelemetry, visit: https://opentelemetry.io/
• For more about OpenTelemetry JavaScript: https://github.com/open-telemetry/opentelemetry-js
• For help or feedback on this project, join us in GitHub Discussions

License

Apache 2.0 - See LICENSE for more information.