feat(telemetry): add stable session identifier headers

packages/datadog-instrumentations/src/child_process.js

@@ -12,7 +12,7 @@ const {
 const childProcessChannel = dc.tracingChannel('datadog:child_process:execution')
 
 // ignored exec method because it calls to execFile directly
-const execAsyncMethods = ['execFile', 'spawn']
+const execAsyncMethods = ['execFile', 'spawn', 'fork']
 
 const names = ['child_process', 'node:child_process']
 
@@ -97,8 +97,10 @@ function wrapChildProcessSyncMethod (returnError, shell = false) {
         return childProcessMethod.apply(this, arguments)
       }
 
-      const childProcessInfo = normalizeArgs(arguments, shell)
+      const callArgs = [...arguments]
+      const childProcessInfo = normalizeArgs(callArgs, shell)
       const context = createContextFromChildProcessInfo(childProcessInfo)
+      context.callArgs = callArgs
 
       return childProcessChannel.start.runStores(context, () => {
         try {
@@ -108,7 +110,7 @@ function wrapChildProcessSyncMethod (returnError, shell = false) {
             return returnError(error, context)
           }
 
-          const result = childProcessMethod.apply(this, arguments)
+          const result = childProcessMethod.apply(this, context.callArgs)
           context.result = result
 
           return result
@@ -131,9 +133,11 @@ function wrapChildProcessCustomPromisifyMethod (customPromisifyMethod, shell) {
       return customPromisifyMethod.apply(this, arguments)
     }
 
-    const childProcessInfo = normalizeArgs(arguments, shell)
+    const callArgs = [...arguments]
+    const childProcessInfo = normalizeArgs(callArgs, shell)
 
     const context = createContextFromChildProcessInfo(childProcessInfo)
+    context.callArgs = callArgs
 
     const { start, end, asyncStart, asyncEnd, error } = childProcessChannel
     start.publish(context)
@@ -143,7 +147,7 @@ function wrapChildProcessCustomPromisifyMethod (customPromisifyMethod, shell) {
       result = Promise.reject(context.abortController.signal.reason || new Error('Aborted'))
     } else {
       try {
-        result = customPromisifyMethod.apply(this, arguments)
+        result = customPromisifyMethod.apply(this, context.callArgs)
       } catch (error) {
         context.error = error
         error.publish(context)
@@ -181,9 +185,11 @@ function wrapChildProcessAsyncMethod (ChildProcess, shell = false) {
         return childProcessMethod.apply(this, arguments)
       }
 
-      const childProcessInfo = normalizeArgs(arguments, shell)
+      const callArgs = [...arguments]
+      const childProcessInfo = normalizeArgs(callArgs, shell)
 
       const context = createContextFromChildProcessInfo(childProcessInfo)
+      context.callArgs = callArgs
       return childProcessChannel.start.runStores(context, () => {
         let childProcess
         if (context.abortController.signal.aborted) {
@@ -194,15 +200,15 @@ function wrapChildProcessAsyncMethod (ChildProcess, shell = false) {
             const error = context.abortController.signal.reason || new Error('Aborted')
             childProcess.emit('error', error)
 
-            const cb = arguments[arguments.length - 1]
+            const cb = context.callArgs[context.callArgs.length - 1]
             if (typeof cb === 'function') {
               cb(error)
             }
 
             childProcess.emit('close')
           })
         } else {
-          childProcess = childProcessMethod.apply(this, arguments)
+          childProcess = childProcessMethod.apply(this, context.callArgs)
         }
 
         if (childProcess) {

packages/datadog-instrumentations/test/child_process.spec.js

@@ -1,6 +1,9 @@
 'use strict'
 
 const assert = require('node:assert/strict')
+const fs = require('node:fs')
+const os = require('node:os')
+const path = require('node:path')
 const { promisify } = require('node:util')
 
 const dc = require('dc-polyfill')
@@ -326,7 +329,6 @@ describe('child process', () => {
                   abortController: sinon.match.instanceOf(AbortController),
                   shell: true,
                 })
-                sinon.assert.calledOnce(start)
                 sinon.assert.calledWithMatch(asyncFinish, {
                   command: 'echo',
                   file: 'echo',
@@ -701,6 +703,130 @@ describe('child process', () => {
           })
         })
       })
+
+      describe('fork', () => {
+        let tmpScript
+
+        before(() => {
+          tmpScript = path.join(os.tmpdir(), `dd-trace-test-fork-${Date.now()}.js`)
+          fs.writeFileSync(tmpScript, 'process.exit(0)')
+        })
+
+        after(() => {
+          try { fs.unlinkSync(tmpScript) } catch (e) { /* ignore */ }
+        })
+
+        it('should execute success callbacks', (done) => {
+          const child = childProcess.fork(tmpScript)
+
+          child.once('close', () => {
+            sinon.assert.calledOnce(start)
+            sinon.assert.calledWithMatch(start, {
+              command: tmpScript,
+              file: tmpScript,
+              shell: false,
+              abortController: sinon.match.instanceOf(AbortController),
+            })
+            sinon.assert.calledOnce(asyncFinish)
+            sinon.assert.calledWithMatch(asyncFinish, {
+              command: tmpScript,
+              file: tmpScript,
+              shell: false,
+              result: 0,
+            })
+            sinon.assert.notCalled(error)
+            done()
+          })
+        })
+
+        it('should publish arguments', (done) => {
+          const child = childProcess.fork(tmpScript, ['--flag'])
+
+          child.once('close', () => {
+            sinon.assert.calledOnce(start)
+            sinon.assert.calledWithMatch(start, {
+              command: `${tmpScript} --flag`,
+              file: tmpScript,
+              fileArgs: ['--flag'],
+              shell: false,
+              abortController: sinon.match.instanceOf(AbortController),
+            })
+            done()
+          })
+        })
+
+        it('should execute error callback for non-existent module', (done) => {
+          const child = childProcess.fork('non_existent_module_test.js')
+
+          child.once('error', () => {})
+
+          child.once('close', () => {
+            sinon.assert.calledOnce(start)
+            sinon.assert.calledWithMatch(start, {
+              command: 'non_existent_module_test.js',
+              file: 'non_existent_module_test.js',
+              shell: false,
+            })
+            sinon.assert.calledOnce(error)
+            done()
+          })
+        })
+      })
+
+      describe('callArgs on context', () => {
+        function injectTestEnv (context) {
+          if (!context.callArgs) return
+          const args = context.callArgs
+          const opts = args[2] != null && typeof args[2] === 'object' ? args[2] : {}
+          args[2] = { ...opts, env: { ...process.env, DD_TEST_VAR: 'injected' } }
+        }
+
+        it('should include callArgs for async methods', (done) => {
+          const child = childProcess.spawn('echo', ['hello'])
+
+          child.once('close', () => {
+            sinon.assert.calledOnce(start)
+            const context = start.firstCall.firstArg
+            assert.ok(Array.isArray(context.callArgs))
+            assert.strictEqual(context.callArgs[0], 'echo')
+            assert.deepStrictEqual(context.callArgs[1], ['hello'])
+            done()
+          })
+        })
+
+        it('should include callArgs for sync methods', () => {
+          childProcess.spawnSync('echo', ['hello'])
+
+          sinon.assert.calledOnce(start)
+          const context = start.firstCall.firstArg
+          assert.ok(Array.isArray(context.callArgs))
+          assert.strictEqual(context.callArgs[0], 'echo')
+          assert.deepStrictEqual(context.callArgs[1], ['hello'])
+        })
+
+        it('should allow subscribers to mutate callArgs for async methods', (done) => {
+          childProcessChannel.subscribe({ start: injectTestEnv })
+
+          const script = 'process.exit(process.env.DD_TEST_VAR === "injected" ? 0 : 1)'
+          const child = childProcess.spawn('node', ['-e', script])
+
+          child.once('close', (code) => {
+            childProcessChannel.unsubscribe({ start: injectTestEnv })
+            assert.strictEqual(code, 0)
+            done()
+          })
+        })
+
+        it('should allow subscribers to mutate callArgs for sync methods', () => {
+          childProcessChannel.subscribe({ start: injectTestEnv })
+
+          const script = 'process.exit(process.env.DD_TEST_VAR === "injected" ? 0 : 1)'
+          const result = childProcess.spawnSync('node', ['-e', script])
+
+          childProcessChannel.unsubscribe({ start: injectTestEnv })
+          assert.strictEqual(result.status, 0)
+        })
+      })
     })
   })
 })

packages/dd-trace/src/config/index.js

@@ -49,6 +49,8 @@ const VALID_PROPAGATION_BEHAVIOR_EXTRACT = new Set(['continue', 'restart', 'igno
 const VALID_LOG_LEVELS = new Set(['debug', 'info', 'warn', 'error'])
 const DEFAULT_OTLP_PORT = 4318
 const RUNTIME_ID = uuid()
+// eslint-disable-next-line eslint-rules/eslint-process-env -- internal propagation, not user config
+const ROOT_SESSION_ID = process.env.DD_ROOT_JS_SESSION_ID || RUNTIME_ID
 const NAMING_VERSIONS = new Set(['v0', 'v1'])
 const DEFAULT_NAMING_VERSION = 'v0'
 
@@ -145,6 +147,8 @@ class Config {
       'runtime-id': RUNTIME_ID,
     })
 
+    this.rootSessionId = ROOT_SESSION_ID
+
     if (this.isCiVisibility) {
       tagger.add(this.tags, {
         [ORIGIN_KEY]: 'ciapp-test',

packages/dd-trace/src/telemetry/send-data.js

@@ -91,12 +91,17 @@ let agentTelemetry = true
  * @returns {Record<string, string>}
  */
 function getHeaders (config, application, reqType) {
+  const sessionId = config.tags['runtime-id']
   const headers = {
     'content-type': 'application/json',
     'dd-telemetry-api-version': 'v2',
     'dd-telemetry-request-type': reqType,
     'dd-client-library-language': application.language_name,
     'dd-client-library-version': application.tracer_version,
+    'dd-session-id': sessionId,
+  }
+  if (config.rootSessionId && config.rootSessionId !== sessionId) {
+    headers['dd-root-session-id'] = config.rootSessionId
   }
   const debug = config.telemetry && config.telemetry.debug
   if (debug) {

packages/dd-trace/src/telemetry/session-propagation.js

@@ -0,0 +1,78 @@
+'use strict'
+
+const dc = require('dc-polyfill')
+
+const childProcessChannel = dc.tracingChannel('datadog:child_process:execution')
+
+let subscribed = false
+let rootSessionId
+let runtimeId
+
+function injectSessionEnv (existingEnv) {
+  // eslint-disable-next-line eslint-rules/eslint-process-env -- not in supported-configurations.json
+  const base = existingEnv == null ? process.env : existingEnv
+  return {
+    ...base,
+    DD_ROOT_JS_SESSION_ID: rootSessionId,
+    DD_PARENT_JS_SESSION_ID: runtimeId,
+  }
+}
+
+function findOptionsIndex (args, shell) {
+  if (Array.isArray(args[1])) {
+    return { index: 2, exists: args[2] != null && typeof args[2] === 'object' }
+  }
+  if (args[1] != null && typeof args[1] === 'object') {
+    return { index: 1, exists: true }
+  }
+  if (!shell && args[2] != null && typeof args[2] === 'object') {
+    return { index: 2, exists: true }
+  }
+  return { index: shell ? 1 : 2, exists: false }
+}
+
+function onChildProcessStart (context) {
+  if (!context.callArgs) return
+
+  const args = context.callArgs
+  const { index, exists } = findOptionsIndex(args, context.shell)
+
+  if (exists) {
+    args[index] = { ...args[index], env: injectSessionEnv(args[index].env) }
+    return
+  }
+
+  const opts = { env: injectSessionEnv(null) }
+
+  if (!context.shell && !Array.isArray(args[1])) {
+    args.splice(1, 0, [])
+  }
+
+  if (typeof args[index] === 'function') {
+    args.splice(index, 0, opts)
+  } else {
+    args[index] = opts
+  }
+}
+
+const handler = { start: onChildProcessStart }
+
+function start (config) {
+  if (!config.telemetry?.enabled || subscribed) return
+  subscribed = true
+
+  rootSessionId = config.rootSessionId
+  runtimeId = config.tags['runtime-id']
+
+  childProcessChannel.subscribe(handler)
+}
+
+function stop () {
+  if (!subscribed) return
+  childProcessChannel.unsubscribe(handler)
+  subscribed = false
+  rootSessionId = undefined
+  runtimeId = undefined
+}
+
+module.exports = { start, stop, _onChildProcessStart: onChildProcessStart }

packages/dd-trace/src/telemetry/telemetry.js

@@ -12,6 +12,7 @@ const endpoints = require('./endpoints')
 const { sendData } = require('./send-data')
 const { manager: metricsManager } = require('./metrics')
 const telemetryLogger = require('./logs')
+const sessionPropagation = require('./session-propagation')
 
 /**
  * @typedef {Record<string, unknown>} TelemetryPayloadObject
@@ -370,6 +371,7 @@ function start (aConfig, thePluginManager) {
   dependencies.start(config, application, host, getRetryData, updateRetryData)
   telemetryLogger.start(config)
   endpoints.start(config, application, host, getRetryData, updateRetryData)
+  sessionPropagation.start(config)
 
   sendData(config, application, host, 'app-started', appStarted(config))
 
@@ -397,6 +399,7 @@ function stop () {
   telemetryStopChannel.publish(getTelemetryData())
 
   endpoints.stop()
+  sessionPropagation.stop()
   config = undefined
 }