feat(tracer): instrument fetch requests (#2293)

* feat(tracer): capture fetch requests * chore: clean up code * chore: swap dummy url in tests * chore: use undici-types * chore: integration tests * tests: update tests * docs: update docs to mention fetch * tests: update integration tests for fetch * improv: handle failed connection case * chore: removed leftover file * Trigger Build --------- Co-authored-by: Alexander Schueren <[email protected]>
aws-powertools · Apr 5, 2024 · cc34400 · cc34400
1 parent 082b626
commit cc34400
Show file tree

Hide file tree

Showing 11 changed files with 548 additions and 20 deletions.
diff --git a/docs/core/tracer.md b/docs/core/tracer.md
@@ -8,7 +8,7 @@ Tracer is an opinionated thin wrapper for [AWS X-Ray SDK for Node.js](https://gi
 ## Key features
 
 * Auto-capturing cold start and service name as annotations, and responses or full exceptions as metadata.
-* Automatically tracing HTTP(S) clients and generating segments for each request.
+* Automatically tracing HTTP(S) clients including `fetch` and generating segments for each request.
 * Supporting tracing functions via decorators, middleware, and manual instrumentation.
 * Supporting tracing AWS SDK v2 and v3 via AWS X-Ray SDK for Node.js.
 * Auto-disable tracing when not running in the Lambda environment.
@@ -211,12 +211,12 @@ If you're looking to shave a few microseconds, or milliseconds depending on your
 
 ### Tracing HTTP requests
 
-When your function makes calls to HTTP APIs, Tracer automatically traces those calls and add the API to the service graph as a downstream service.
+When your function makes outgoing requests to APIs, Tracer automatically traces those calls and adds the API to the service graph as a downstream service.
 
 You can opt-out from this feature by setting the **`POWERTOOLS_TRACER_CAPTURE_HTTPS_REQUESTS=false`** environment variable or by passing the `captureHTTPSRequests: false` option to the `Tracer` constructor.
 
 !!! info
-    The following snippet shows how to trace [axios](https://www.npmjs.com/package/axios) requests, but you can use any HTTP client library built on top of [http](https://nodejs.org/api/http.html) or [https](https://nodejs.org/api/https.html).
+    The following snippet shows how to trace [`fetch`](https://developer.mozilla.org/en-US/docs/Web/API/fetch) requests, but you can use any HTTP client library built on top it, or on [http](https://nodejs.org/api/http.html), and [https](https://nodejs.org/api/https.html).
     Support to 3rd party HTTP clients is provided on a best effort basis.
 
 === "index.ts"
@@ -225,9 +225,6 @@ You can opt-out from this feature by setting the **`POWERTOOLS_TRACER_CAPTURE_HT
     --8<-- "docs/snippets/tracer/captureHTTP.ts"
     ```
 
-    1.  You can install the [axios](https://www.npmjs.com/package/axios) package using `npm i axios`
-=== "Example Raw X-Ray Trace excerpt"
-
     ```json hl_lines="6 9 12-21"
     {
         "id": "22883fbc730e3a0b",

diff --git a/docs/snippets/tracer/captureHTTP.ts b/docs/snippets/tracer/captureHTTP.ts
@@ -1,11 +1,10 @@
 import { Tracer } from '@aws-lambda-powertools/tracer';
-import axios from 'axios'; // (1)
 
 new Tracer({ serviceName: 'serverlessAirline' });
 
 export const handler = async (
   _event: unknown,
   _context: unknown
 ): Promise<void> => {
-  await axios.get('https://httpbin.org/status/200');
+  await fetch('https://httpbin.org/status/200');
 };
diff --git a/packages/tracer/src/Tracer.ts b/packages/tracer/src/Tracer.ts
@@ -16,7 +16,7 @@ import type {
   CaptureMethodOptions,
 } from './types/Tracer.js';
 import { ProviderService } from './provider/ProviderService.js';
-import type { ProviderServiceInterface } from './types/ProviderServiceInterface.js';
+import type { ProviderServiceInterface } from './types/ProviderService.js';
 import type { Segment, Subsegment } from 'aws-xray-sdk-core';
 import xraySdk from 'aws-xray-sdk-core';
 const { Subsegment: XraySubsegment } = xraySdk;
@@ -153,6 +153,7 @@ class Tracer extends Utility implements TracerInterface {
     this.provider = new ProviderService();
     if (this.isTracingEnabled() && this.captureHTTPsRequests) {
       this.provider.captureHTTPsGlobal();
+      this.provider.instrumentFetch();
     }
     if (!this.isTracingEnabled()) {
       // Tell x-ray-sdk to not throw an error if context is missing but tracing is disabled

diff --git a/packages/tracer/src/provider/ProviderService.ts b/packages/tracer/src/provider/ProviderService.ts
@@ -1,8 +1,9 @@
-import { Namespace } from 'cls-hooked';
+import type { Namespace } from 'cls-hooked';
 import type {
   ProviderServiceInterface,
   ContextMissingStrategy,
-} from '../types/ProviderServiceInterface.js';
+  HttpSubsegment,
+} from '../types/ProviderService.js';
 import type { Segment, Subsegment, Logger } from 'aws-xray-sdk-core';
 import xraySdk from 'aws-xray-sdk-core';
 const {
@@ -21,6 +22,13 @@ const {
   setLogger,
 } = xraySdk;
 import { addUserAgentMiddleware } from '@aws-lambda-powertools/commons';
+import { subscribe } from 'node:diagnostics_channel';
+import {
+  findHeaderAndDecode,
+  getOriginURL,
+  isHttpSubsegment,
+} from './utilities.js';
+import type { DiagnosticsChannel } from 'undici-types';
 
 class ProviderService implements ProviderServiceInterface {
   public captureAWS<T>(awssdk: T): T {
@@ -70,6 +78,119 @@ class ProviderService implements ProviderServiceInterface {
     return getSegment();
   }
 
+  /**
+   * Instrument `fetch` requests with AWS X-Ray
+   *
+   * The instrumentation is done by subscribing to the `undici` events. When a request is created,
+   * a new subsegment is created with the hostname of the request.
+   *
+   * Then, when the headers are received, the subsegment is updated with the request and response details.
+   *
+   * Finally, when the request is completed, the subsegment is closed.
+   *
+   * @see {@link https://nodejs.org/api/diagnostics_channel.html#diagnostics_channel_channel_publish | Diagnostics Channel - Node.js Documentation}
+   */
+  public instrumentFetch(): void {
+    /**
+     * Create a segment at the start of a request made with `undici` or `fetch`.
+     *
+     * @note that `message` must be `unknown` because that's the type expected by `subscribe`
+     *
+     * @param message The message received from the `undici` channel
+     */
+    const onRequestStart = (message: unknown): void => {
+      const { request } = message as DiagnosticsChannel.RequestCreateMessage;
+
+      const parentSubsegment = this.getSegment();
+      if (parentSubsegment && request.origin) {
+        const origin = getOriginURL(request.origin);
+        const method = request.method;
+
+        const subsegment = parentSubsegment.addNewSubsegment(origin.hostname);
+        subsegment.addAttribute('namespace', 'remote');
+
+        (subsegment as HttpSubsegment).http = {
+          request: {
+            url: origin.hostname,
+            method,
+          },
+        };
+
+        this.setSegment(subsegment);
+      }
+    };
+
+    /**
+     * Enrich the subsegment with the response details, and close it.
+     * Then, set the parent segment as the active segment.
+     *
+     * @note that `message` must be `unknown` because that's the type expected by `subscribe`
+     *
+     * @param message The message received from the `undici` channel
+     */
+    const onResponse = (message: unknown): void => {
+      const { response } = message as DiagnosticsChannel.RequestHeadersMessage;
+
+      const subsegment = this.getSegment();
+      if (isHttpSubsegment(subsegment)) {
+        const status = response.statusCode;
+        const contentLenght = findHeaderAndDecode(
+          response.headers,
+          'content-length'
+        );
+
+        subsegment.http = {
+          ...subsegment.http,
+          response: {
+            status,
+            ...(contentLenght && {
+              content_length: parseInt(contentLenght),
+            }),
+          },
+        };
+
+        if (status === 429) {
+          subsegment.addThrottleFlag();
+        }
+        if (status >= 400 && status < 500) {
+          subsegment.addErrorFlag();
+        } else if (status >= 500 && status < 600) {
+          subsegment.addFaultFlag();
+        }
+
+        subsegment.close();
+        this.setSegment(subsegment.parent);
+      }
+    };
+
+    /**
+     * Add an error to the subsegment when the request fails.
+     *
+     * This is used to handle the case when the request fails to establish a connection with the server or timeouts.
+     * In all other cases, for example, when the server returns a 4xx or 5xx status code, the error is added in the `onResponse` function.
+     *
+     * @note that `message` must be `unknown` because that's the type expected by `subscribe`
+     *
+     * @param message The message received from the `undici` channel
+     */
+    const onError = (message: unknown): void => {
+      const { error } = message as DiagnosticsChannel.RequestErrorMessage;
+
+      const subsegment = this.getSegment();
+      if (isHttpSubsegment(subsegment)) {
+        subsegment.addErrorFlag();
+        error instanceof Error && subsegment.addError(error, true);
+
+        subsegment.close();
+        this.setSegment(subsegment.parent);
+      }
+    };
+
+    subscribe('undici:request:create', onRequestStart);
+    subscribe('undici:request:headers', onResponse);
+    subscribe('undici:request:error', onError);
+  }
+
   public putAnnotation(key: string, value: string | number | boolean): void {
     const segment = this.getSegment();
     if (segment === undefined) {

diff --git a/packages/tracer/src/provider/utilities.ts b/packages/tracer/src/provider/utilities.ts
@@ -0,0 +1,63 @@
+import type { HttpSubsegment } from '../types/ProviderService.js';
+import type { Segment, Subsegment } from 'aws-xray-sdk-core';
+import { URL } from 'node:url';
+
+const decoder = new TextDecoder();
+
+/**
+ * The `fetch` implementation based on `undici` includes the headers as an array of encoded key-value pairs.
+ * This function finds the header with the given key and decodes the value.
+ *
+ * The function walks through the array of encoded headers and decodes the key of each pair.
+ * If the key matches the given key, the function returns the decoded value of the next element in the array.
+ *
+ * @param encodedHeaders The array of encoded headers
+ * @param key The key to search for
+ */
+const findHeaderAndDecode = (
+  encodedHeaders: Uint8Array[],
+  key: string
+): string | null => {
+  let foundIndex = -1;
+  for (let i = 0; i < encodedHeaders.length; i += 2) {
+    const header = decoder.decode(encodedHeaders[i]);
+    if (header.toLowerCase() === key) {
+      foundIndex = i;
+      break;
+    }
+  }
+
+  if (foundIndex === -1) {
+    return null;
+  }
+
+  return decoder.decode(encodedHeaders[foundIndex + 1]);
+};
+
+/**
+ * Type guard to check if the given subsegment is an `HttpSubsegment`
+ *
+ * @param subsegment The subsegment to check
+ */
+const isHttpSubsegment = (
+  subsegment: Segment | Subsegment | undefined
+): subsegment is HttpSubsegment => {
+  return (
+    subsegment !== undefined &&
+    'http' in subsegment &&
+    'parent' in subsegment &&
+    'namespace' in subsegment &&
+    subsegment.namespace === 'remote'
+  );
+};
+
+/**
+ * Convert the origin url to a URL object when it is a string
+ *
+ * @param origin The origin url
+ */
+const getOriginURL = (origin: string | URL): URL => {
+  return origin instanceof URL ? origin : new URL(origin);
+};
+
+export { findHeaderAndDecode, isHttpSubsegment, getOriginURL };
diff --git a/...cer/src/types/ProviderServiceInterface.ts → packages/tracer/src/types/ProviderService.ts b/...cer/src/types/ProviderServiceInterface.ts → packages/tracer/src/types/ProviderService.ts
@@ -40,9 +40,35 @@ interface ProviderServiceInterface {
 
   captureHTTPsGlobal(): void;
 
+  /**
+   * Instrument `fetch` requests with AWS X-Ray
+   */
+  instrumentFetch(): void;
+
   putAnnotation(key: string, value: string | number | boolean): void;
 
   putMetadata(key: string, value: unknown, namespace?: string): void;
 }
 
-export type { ProviderServiceInterface, ContextMissingStrategy };
+/**
+ * Subsegment that contains information for a request made to a remote service
+ */
+interface HttpSubsegment extends Subsegment {
+  namespace: 'remote';
+  http: {
+    request?: {
+      url: string;
+      method?: string;
+    };
+    response?: {
+      status: number;
+      content_length?: number;
+    };
+  };
+}
+
+export type {
+  ProviderServiceInterface,
+  ContextMissingStrategy,
+  HttpSubsegment,
+};
diff --git a/packages/tracer/src/types/Tracer.ts b/packages/tracer/src/types/Tracer.ts
@@ -22,6 +22,9 @@ import type { Segment, Subsegment } from 'aws-xray-sdk-core';
 type TracerOptions = {
   enabled?: boolean;
   serviceName?: string;
+  /**
+   * Whether to trace outgoing HTTP requests made with the `http`, `https`, or `fetch` modules
+   */
   captureHTTPsRequests?: boolean;
   customConfigService?: ConfigServiceInterface;
 };

diff --git a/packages/tracer/tests/e2e/asyncHandler.decorator.test.functionCode.ts b/packages/tracer/tests/e2e/asyncHandler.decorator.test.functionCode.ts
@@ -55,10 +55,13 @@ export class MyFunctionBase {
           Item: { id: `${serviceName}-${event.invocation}-sdkv3` },
         })
       );
-      await axios.get(
-        'https://docs.powertools.aws.dev/lambda/typescript/latest/',
-        { timeout: 5000 }
-      );
+      const url = 'https://docs.powertools.aws.dev/lambda/typescript/latest/';
+      // Add conditional behavior because fetch is not available in Node.js 16 - this can be removed once we drop support for Node.js 16
+      if (process.version.startsWith('v16')) {
+        await axios.get(url, { timeout: 5000 });
+      } else {
+        await fetch(url);
+      }
 
       const res = this.myMethod();
       if (event.throw) {