Node.js中Axios调用生成的OpenTelemetry Span无法与父Span关联

环境与依赖 #

• Node.js v18.16.1
• TypeScript
• 核心依赖：

"@opentelemetry/api": "^1.3.0",
  "@opentelemetry/auto-instrumentations-node": "^0.38.0",
  "@opentelemetry/context-async-hooks": "^1.15.0",
  "@opentelemetry/instrumentation": "^0.41.0",
  "@opentelemetry/instrumentation-http": "^0.35.1",
  "@opentelemetry/resources": "^1.9.1",
  "@opentelemetry/sdk-metrics": "^1.9.1",
  "@opentelemetry/sdk-node": "^0.35.1",
  "@opentelemetry/sdk-trace-base": "^1.9.1",
  "@opentelemetry/sdk-trace-node": "^1.9.1",
  "@opentelemetry/semantic-conventions": "^1.9.1",
  "axios": "^1.4.0"

配置代码 #

otel/tracing.js #

const opentelemetry = require('@opentelemetry/api')
const { Resource } = require("@opentelemetry/resources");
const { SemanticResourceAttributes } = require("@opentelemetry/semantic-conventions");
const { NodeTracerProvider } = require("@opentelemetry/sdk-trace-node");
const { registerInstrumentations } = require("@opentelemetry/instrumentation");
const { ConsoleSpanExporter, SimpleSpanProcessor} = require("@opentelemetry/sdk-trace-base");
const { AsyncHooksContextManager } = require('@opentelemetry/context-async-hooks');
const {
    HttpInstrumentation,
} = require("@opentelemetry/instrumentation-http");


const contextManager = new AsyncHooksContextManager();

contextManager.enable()

opentelemetry.context.setGlobalContextManager(contextManager)


registerInstrumentations({
    instrumentations: [new HttpInstrumentation()],
});

const resource =
    Resource.default().merge(
        new Resource({
            [SemanticResourceAttributes.SERVICE_NAME]: 'poc-otel-tracing',
            [SemanticResourceAttributes.SERVICE_VERSION]: "0.1.0",
        })
    );

const provider = new NodeTracerProvider({
    resource: resource
});

const exporter = new ConsoleSpanExporter();
const processor = new SimpleSpanProcessor(exporter);


provider.addSpanProcessor(processor);

provider.register();

// process.on('exit', function() {
//     void exporter.shutdown()
// });

src/index.ts #

import axios from "axios";

import opentelemetry, {Span} from "@opentelemetry/api";
const tracer = opentelemetry.trace.getTracer('poc-otel-tracing');

const API_BASE_URL = 'https://jsonplaceholder.typicode.com/posts';

const func = async() => {

    const rootSpan = tracer.startSpan('root-span')
    await apiCall(rootSpan)
    rootSpan.end()
}

const apiCall = async (parentSpan: Span) => {

    const ctx = opentelemetry.trace.setSpan(opentelemetry.context.active(), parentSpan)

    const childSpan = tracer.startSpan('api-call', undefined, ctx)

    try {
        await axios.get(API_BASE_URL);
    } catch (e) {
        childSpan.recordException(e);
    }

    childSpan.end()
}

void func()

运行命令 #

npm run build && node --require "./otel/tracing.js" dist/index.js

问题现象 #

控制台输出的Span中，Axios生成的HTTPS GET Span拥有独立的traceId，未与api-call Span形成父子关联：

{
  traceId: '15b261f7332feb1fdc79cff5a280529b',
  parentId: undefined,
  traceState: undefined,
  name: 'HTTPS GET',
  id: '477643c3a9536749',
  kind: 2,
  timestamp: 1690206903281000,
  duration: 128485,
  attributes: {
    'http.url': 'https://jsonplaceholder.typicode.com/posts',
    'http.method': 'GET',
    'http.target': '/posts',
    'net.peer.name': 'jsonplaceholder.typicode.com',
    'http.host': 'jsonplaceholder.typicode.com:443',
    'net.peer.ip': '172.64.162.6',
    'net.peer.port': 443,
    'http.status_code': 200,
    'http.status_text': 'OK',
    'http.flavor': '1.1',
    'net.transport': 'ip_tcp'
  },
  status: { code: 0 },
  events: [],
  links: []
}
{
  traceId: 'ba35dbc454728c2bbc5b30a435dec3bb',
  parentId: 'a3897743fb82276c',
  traceState: undefined,
  name: 'api-call',
  id: 'ec3d00cb4ad95d26',
  kind: 0,
  timestamp: 1690206903278000,
  duration: 136158,
  attributes: {},
  status: { code: 0 },
  events: [],
  links: []
}
{
  traceId: 'ba35dbc454728c2bbc5b30a435dec3bb',
  parentId: undefined,
  traceState: undefined,
  name: 'root-span',
  id: 'a3897743fb82276c',
  kind: 0,
  timestamp: 1690206903278000,
  duration: 136633,
  attributes: {},
  status: { code: 0 },
  events: [],
  links: []
}

预期结果 #

HTTPS GET Span应继承api-call Span的上下文，属于同一Trace，且parentId指向api-call的Span ID：

{
  traceId: 'ba35dbc454728c2bbc5b30a435dec3bb',
  parentId: 'ec3d00cb4ad95d26',
  traceState: undefined,
  name: 'HTTPS GET',
  id: '477643c3a9536749',
  kind: 2,
  timestamp: 1690206903281000,
  duration: 128485,
  attributes: {
    'http.url': 'https://jsonplaceholder.typicode.com/posts',
    'http.method': 'GET',
    'http.target': '/posts',
    'net.peer.name': 'jsonplaceholder.typicode.com',
    'http.host': 'jsonplaceholder.typicode.com:443',
    'net.peer.ip': '172.64.162.6',
    'net.peer.port': 443,
    'http.status_code': 200,
    'http.status_text': 'OK',
    'http.flavor': '1.1',
    'net.transport': 'ip_tcp'
  },
  status: { code: 0 },
  events: [],
  links: []
}
{
  traceId: 'ba35dbc454728c2bbc5b30a435dec3bb',
  parentId: 'a3897743fb82276c',
  traceState: undefined,
  name: 'api-call',
  id: 'ec3d00cb4ad95d26',
  kind: 0,
  timestamp: 1690206903278000,
  duration: 136158,
  attributes: {},
  status: { code: 0 },
  events: [],
  links: []
}
{
  traceId: 'ba35dbc454728c2bbc5b30a435dec3bb',
  parentId: undefined,
  traceState: undefined,
  name: 'root-span',
  id: 'a3897743fb82276c',
  kind: 0,
  timestamp: 1690206903278000,
  duration: 136633,
  attributes: {},
  status: { code: 0 },
  events: [],
  links: []
}

问题原因与解决方法 #

原因 #

代码中仅创建了childSpan，但未将其上下文绑定到Axios调用的异步执行流中。OpenTelemetry的AsyncHooksContextManager需要显式将Span上下文关联到当前操作，否则HTTP instrumentation无法识别当前Span，会生成独立的Trace。

解决方法 #

有两种常用方式将Span上下文绑定到异步操作：

方法1：使用tracer.startActiveSpan（推荐）

startActiveSpan会自动将创建的Span设置为当前上下文的活跃Span，无需手动绑定：

const apiCall = async (parentSpan: Span) => {
    const ctx = opentelemetry.trace.setSpan(opentelemetry.context.active(), parentSpan);
    
    await tracer.startActiveSpan('api-call', {}, ctx, async (childSpan) => {
        try {
            await axios.get(API_BASE_URL);
        } catch (e) {
            childSpan.recordException(e);
        } finally {
            childSpan.end();
        }
    });
}

方法2：使用opentelemetry.context.with

手动将Span上下文绑定到Axios调用的执行上下文：

const apiCall = async (parentSpan: Span) => {
    const parentCtx = opentelemetry.trace.setSpan(opentelemetry.context.active(), parentSpan);
    const childSpan = tracer.startSpan('api-call', undefined, parentCtx);
    // 创建包含childSpan的上下文
    const childCtx = opentelemetry.trace.setSpan(parentCtx, childSpan);
    
    try {
        // 在childCtx的上下文中执行Axios调用
        await opentelemetry.context.with(childCtx, async () => {
            await axios.get(API_BASE_URL);
        });
    } catch (e) {
        childSpan.recordException(e);
    } finally {
        childSpan.end();
    }
}

修改后重新运行，Axios生成的HTTPS GET Span会正确关联api-call Span，属于同一Trace。

内容的提问来源于stack exchange，提问作者Liviu Gheorghe