OpenTelemetry for JavaScript on Fastly Compute

An implementation of the OpenTelemetry JavaScript API for Fastly Compute.

Generate traces like this one to follow activity and time in your Compute applications:

index.js:

/// <reference types="@fastly/js-compute" />

import './telemetry.js'
import { context, trace } from "@opentelemetry/api";

addEventListener("fetch", (event) => event.respondWith(handleRequest(event)));
async function handleRequest(event) {
  const tracer = trace.getTracerProvider()
    .getTracer('my-tracer');

  const mySpan = tracer.startSpan('my-task');
  context.with(trace.setSpan(context.active(), mySpan), () => {
    doTask();
  });
  mySpan.end();

  return new Response('OK', {
    status: 200,
    headers: new Headers({"Content-Type": "text/plain"}),
  });
}

telemetry.js:

import { context, trace } from "@opentelemetry/api";
import { Resource } from "@opentelemetry/resources";
import { SEMRESATTRS_SERVICE_NAME } from "@opentelemetry/semantic-conventions";

import { FastlySDK } from "@fastly/compute-js-opentelemetry/sdk-fastly";
import { OTLPTraceExporter } from "@fastly/compute-js-opentelemetry/exporter-trace-otlp-fastly-backend";
import { getComputeJsAutoInstrumentations } from "@fastly/compute-js-opentelemetry/auto-instrumentations-compute-js";

const sdk = new FastlySDK({
  traceExporter: new OTLPTraceExporter({ backend: 'otlp-collector' }),
  instrumentations: [ getComputeJsAutoInstrumentations(), ],
  resource: new Resource({ [SEMRESATTRS_SERVICE_NAME]: 'readme-demo', }),
});
await sdk.start();

This implementation extends the standard interfaces and objects provided by the OpenTelemetry JavaScript API and SDK, adapting them for use on the Fastly Compute platform.

New!

As of v0.3.0, @fastly/compute-js-opentelemetry no longer requires Webpack.

Additionally, if you're using the FastlySDK module, you will need @fastly/js-compute v3.11.0 or newer and enable Top Level Await. Find where you call the binary js-compute-runtime and add the flag like so js-compute-runtime --experimental-enable-top-level-await (This is usually in either package.json or fastly.toml).

Look at the demos included in this package for working examples.

Modules

Whereas opentelemetry-js would separate each concern into its own npm package, we provide our components as a single package with multiple exports.

Module (links to documentation)	Export Name	Description
Fastly OpenTelemetry SDK	@fastly/compute-js-opentelemetry/opentelemetry-sdk-fastly	A utility library that simplifies the initialization and coordination of OpenTelemetry components in use with a Compute JavaScript application.
Fastly Compute JavaScript Lifecycle instrumentation	@fastly/compute-js-opentelemetry/opentelemetry-instrumentation-fastly-compute-js	An instrumentation library that generates traces for the Compute application lifecycle.
Fastly Backend Fetch instrumentation	@fastly/compute-js-opentelemetry/opentelemetry-instrumentation-fastly-backend-fetch	An instrumentation library that generates traces for fetch() calls to Fastly backends.
Fastly Compute JavaScript auto-instrumentations	@fastly/compute-js-opentelemetry/auto-instrumentations-compute-js	A meta package that automatically loads instrumentations for Compute JavaScript.
Trace SDK for Fastly	@fastly/compute-js-opentelemetry/opentelemetry-sdk-trace-fastly	Provides Span Processor, Tracer Provider, and Context Manager implementations for use with a Compute JavaScript application.
Trace Exporter for Fastly backend	@fastly/compute-js-opentelemetry/exporter-trace-otlp-fastly-backend	A Span Exporter implementation that exports traces using the OTLP/HTTP JSON format over a Fastly backend.
Trace Exporter for Fastly named log provider	@fastly/compute-js-opentelemetry/exporter-trace-otlp-fastly-logger	A Span Exporter implementation that exports traces using the OTLP/HTTP JSON format over a Fastly named log provider.
Metrics SDK for Fastly	@fastly/compute-js-opentelemetry/opentelemetry-sdk-metrics-fastly	Provides a Metric Reader implementation for use with a Compute JavaScript application.
Metrics Exporter for Fastly backend	@fastly/compute-js-opentelemetry/exporter-metrics-otlp-fastly-backend	A Metric Exporter implementation that exports metrics using the OTLP/HTTP JSON format over a Fastly backend.
Metrics Exporter for Fastly named log provider	@fastly/compute-js-opentelemetry/exporter-metrics-otlp-fastly-logger	A Metric Exporter implementation that exports metrics using the OTLP/HTTP JSON format over a Fastly named log provider.
Exporter Base Classes	@fastly/compute-js-opentelemetry/otlp-exporter-fastly-base	Base classes for exporters, containing common code used by the exporter classes listed above.
Diagnostic Logger for Fastly named log provider	@fastly/compute-js-opentelemetry/diag-fastly-logger	A DiagLogger implementation that logs to a Fastly named log provider.
Core Utilities (Internal Use)	(Not exported)	Utilities intended to be used internally by the library and not from user code.

Examples

See the examples in the /examples directory.

Example Directory	Description
readme-demo	Example demo from the beginning of this README
readme-demo-ts	Example demo from the beginning of this README, ported to TypeScript
basic-tracing-example	Basic Tracing Example
basic-tracing-example-ts	Basic Tracing Example in TypeScript
basic-metrics-example	Basic Metrics Example
otel-demo	Example that demonstrates OpenTelemetry traces that start at the Edge and nest into an operation at the backend.
otel-http-proxy	A sample application designed to collect traces as an HTTPS log endpoint for a Fastly service, sending them to an OpenTelemetry collector.

Use with TypeScript

This library is designed to work with both JavaScript and TypeScript applications written for Fastly Compute.

Check out readme-demo-ts and basic-tracing-example-ts in the examples directory for working sample programs using TypeScript.

Notes

Environment Variables

OpenTelemetry defines a well-documented set of environment variables that are designed to allow you to configure defaults for its libraries. However, JavaScript applications built to run on Fastly's Compute platform perform their upfront initialization at build time ("build-time initialization") rather than during each invocation (that's how they start up so fast). This means that the platform cannot provide a way to read from the environment during build-time initialization, when most OpenTelemetry libraries are initializing.

Therefore, when using this library, the getEnv() function will always return the default values defined by OpenTelemetry. Any changes you require to these defaults need to be made programmatically.

Compatibility

Some opentelemetry-js modules are not currently compatible with Fastly Compute. The table below is a non-comprehensive list of such components.

Component	Package	Reason / Workaround
OTLPTraceExporter	@opentelemetry/exporter-trace-otlp-http	Relies on XmlHttpRequest and navigator.beacon, which are not available in Compute. Use OTLPTraceExporter from @fastly/compute-js-opentelemetry/exporter-trace-otlp-fastly-backend or @fastly/compute-js-opentelemetry/exporter-trace-otlp-fastly-logger instead.
OTLPMetricExporter	@opentelemetry/exporter-metrics-otlp-http	Relies on XmlHttpRequest and navigator.beacon, which are not available in Compute. Use OTLPTraceExporter from @fastly/compute-js-opentelemetry/exporter-metrics-otlp-fastly-backend or @fastly/compute-js-opentelemetry/exporter-metrics-otlp-fastly-logger instead.
ZoneContextManager	@opentelemetry/context-zone @opentelemetry/context-zone-peer-dep	Relies on zone.js, which is incompatible with Compute. Use FastlyStackContextManager.
AsyncHooksContextManager AsyncLocalStorageContextManager	@opentelemetry/context-async-hooks	Relies on async_hooks, which is not available in Compute. Use FastlyStackContextManager.
NodeSDK	@opentelemetry/sdk-node	This SDK is for Node, which is not compatible with Compute. Use FastlySDK.
Instrumentations included in opentelemetry-js	@opentelemetry/instrumentation-* @opentelemetry/auto-instrumentations-node @opentelemetry/auto-instrumentations-web	These rely on other frameworks and modules that are not compatible with Compute. Use FastlyComputeJsInstrumentation and FastlyBackendFetchInstrumentation.

Issues

If you encounter any non-security-related bug or unexpected behavior, please file an issue using the bug report template.

Security issues

Please see our SECURITY.md for guidance on reporting security-related issues.

License

MIT.