OpenTelemetry for JavaScript on Compute@Edge
An implementation of the OpenTelemetry JavaScript API for
Fastly Compute@Edge.
Generate traces like this one to follow activity and time in your Compute@Edge applications:

index.js:
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 { SemanticResourceAttributes } 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({ [SemanticResourceAttributes.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@Edge platform.
Whereas opentelemetry-js
would separate each concern into its own npm
package,
we provide our components as a single package with multiple exports.
Modules
The table below lists the modules included in this package.
Examples
See the examples in the /examples
directory.
readme-demo | Example demo from the beginning of this README |
basic-tracing-example | Basic Tracing Example |
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. |
Webpack
Compute@Edge JavaScript applications can be
compiled as a web worker using Webpack as
part of their build process, which is useful for applying techniques such as using polyfills and shims.
In order to use the OpenTelemetry packages that we rely on, additions need to be made to the webpack.config.js
configuration file. These changes are included in a helper module, @fastly/compute-js-opentelemetry/webpack-helpers
,
so that they may be applied as such:
const webpackHelpers = require("@fastly/compute-js-opentelemetry/webpack-helpers");
module.exports = {
entry: "./src/index.js",
};
module.exports = webpackHelpers.apply(module.exports);
The use of this module is technically optional, but if you do choose not to use it, you will have to
make the appropriate modifications yourself. See webpack-helpers for
details.
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@Edge 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@Edge.
The table below is a non-comprehensive list of such components.
BatchSpanProcessor | @opentelemetry/sdk-trace-base | Relies on setTimeout . Use FastlySpanProcessor . |
NodeSDK | @opentelemetry/sdk-node | Relies on BatchSpanProcessor as well as platforms detectors that are incompatible with Compute@Edge. Use FastlySDK . |
OTLPTraceExporter | @opentelemetry/exporter-trace-otlp-http | Relies on http and https , which are not available in Compute@Edge. 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 http and https , which are not available in Compute@Edge. 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@Edge. Use FastlyStackContextManager . |
AsyncHooksContextManager
AsyncLocalStorageContextManager | @opentelemetry/context-async-hooks | Relies on async_hooks , which is not available in Compute@Edge. Use FastlyStackContextManager . |
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@Edge. 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.