升级@opentelemetry/exporter-trace-otlp-grpc后出现连接未建立错误

原本运行正常的TypeScript服务，可将指标与追踪数据导出到OpenTelemetry Docker容器，升级至最新API后出现连接错误，问题集中在新API方法PeriodicExportingMetricReader。

OTel Collector配置 #

receivers:
    otlp:
        protocols:
            grpc:

exporters:
    prometheus:
        endpoint: '0.0.0.0:8889'
        namespace: B2B
        metric_expiration: 525600m

    jaeger:
        endpoint: jaeger:14250
        tls:
            insecure: true

processors:
    batch:

extensions:
    health_check:
    pprof:
        endpoint: :1888
    zpages:
        endpoint: :55679

service:
    extensions: [pprof, zpages, health_check]
    telemetry:
        logs:
            level: "info"
            development: true
            encoding: "json"
    pipelines:
        traces:
            receivers: [otlp]
            processors: [batch]
            exporters: [jaeger]
        metrics:
            receivers: [otlp]
            exporters: [prometheus]

项目包依赖配置 #

{
    "@opentelemetry/api": "^1.3.0",
    "@opentelemetry/api-metrics": "^0.33.0",
    "@opentelemetry/host-metrics": "^0.31.0",
    "@opentelemetry/sdk-metrics": "^1.8.0",
    "@opentelemetry/sdk-trace-node": "^1.8.0",
    "@opentelemetry/exporter-metrics-otlp-grpc": "^0.35.1"
}

核心实现代码 #

export async function startTracing() {
  try {
    const collectorOptions = {
      url: `grpc://localhost:4317`,
    };

    const traceExporter = new OTLPTraceExporter(collectorOptions);

    const metricExporter = new PeriodicExportingMetricReader({
      exporter: new OTLPMetricExporter(collectorOptions),
      exportIntervalMillis: 5000,
    });

    const sdk = new NodeSDK({
      resource: new Resource(monitoringAttributes),
      traceExporter: traceExporter,
      metricReader: metricExporter,
      spanProcessor: new BatchSpanProcessor(traceExporter, {
        scheduledDelayMillis: 500,
      }),
      instrumentations: [
        new HttpInstrumentation(),
        new ExpressInstrumentation(),
      ],
    });

    await sdk
      .start()
      .then(() => console.log('Tracing initialized'))
      .catch((error) => console.log('Error initializing tracing', error));

    // gracefully shut down the SDK on process exit
    process.on('SIGTERM', () => {
      sdk
        .shutdown()
        .then(() => console.log('Tracing terminated'))
        .catch((error) => console.log('Error terminating tracing', error))
        .finally(() => process.exit(0));
    });
  } catch (e) {
    console.log('startTracing. error occurred', e);
  }
}

错误日志 #

2023-03-01 16:51:13 Service request {
2023-03-01 16:51:13   resourceMetrics: [
2023-03-01 16:51:13     { resource: [Object], schemaUrl: undefined, scopeMetrics: [Array] }
2023-03-01 16:51:13   ]
2023-03-01 16:51:13 }
2023-03-01 16:51:13 {"stack":"Error: PeriodicExportingMetricReader: metrics export failed (error Error: 14 UNAVAILABLE: No connection established)\n    at PeriodicExportingMetricReader._doRun (/app/node_modules/@opentelemetry/sdk-metrics/src/export/PeriodicExportingMetricReader.ts:122:13)\n    at processTicksAndRejections (internal/process/task_queues.js:95:5)\n    at PeriodicExportingMetricReader._runOnce (/app/node_modules/@opentelemetry/sdk-metrics/src/export/PeriodicExportingMetricReader.ts:94:7)","message":"PeriodicExportingMetricReader: metrics export failed (error Error: 14 UNAVAILABLE: No connection established)","name":"Error"}

排查线索 #

• 版本兼容性问题：当前sdk-metrics（1.8.0）和exporter-metrics-otlp-grpc（0.35.1）版本不匹配。OpenTelemetry的SDK与exporter版本需严格对齐，建议将所有OpenTelemetry相关包统一升级到同一主版本（如1.x系列，0.35.x属于旧版本，与1.x不兼容）。
• 网络连接问题：
  • 若服务运行在Docker容器中，localhost:4317指向容器内部而非宿主机，需替换为Docker服务名称（如otel-collector:4317，前提是OTel Collector在同一Docker网络）或宿主机IP。
  • 检查OTel Collector的OTLP gRPC端口（4317）是否正常监听，可通过docker exec <collector-container> netstat -tulpn验证端口状态。
• Exporter配置格式：新版本OTLPMetricExporter可能要求使用endpoint字段而非url，同时显式指定不安全连接（未启用TLS时）：

  const collectorOptions = {
    endpoint: 'otel-collector:4317',
    credentials: grpc.credentials.createInsecure(),
  };
  

• 权限与防火墙：检查服务容器是否有权限访问OTel Collector端口，宿主机或Docker网络是否有防火墙规则拦截4317端口的gRPC流量。
• 初始化校验：确认PeriodicExportingMetricReader是否正确关联到SDK的metricReader字段，当前写法正确，但需排查是否存在重复初始化或配置冲突。

内容的提问来源于stack exchange，提问作者Laurence Fass