链路追踪 - 函数计算

本文介绍Java 11运行环境的链路追踪相关内容。

背景信息

阿里云链路追踪服务（Tracing Analysis）基于OpenTracing标准，兼容开源社区，为分布式应用的开发者提供了完整地分布式调用链查询和诊断、分布式拓扑动态发现、应用性能实时汇总等功能。

函数计算与链路追踪集成后，支持使用Jaeger SDK和OpenTelemetry上传链路信息，使您能够跟踪函数的执行，帮助您快速分析和诊断Serverless架构下的性能瓶颈，提高Serverless场景的开发诊断效率。

功能简介

您可以在函数计算控制台配置链路追踪。具体操作，请参见配置链路追踪。

为服务开启链路追踪后，函数计算会自动记录请求在系统侧的耗时，包含冷启动耗时、Initializer函数的耗时和函数的执行时间等。关于下图中系统Span的说明，请参见Span名称说明。

如您还需查看函数内业务侧的耗时，例如，在函数内访问RDS，NAS等服务的耗时，可以通过创建自定义Span来实现。

示例代码

函数计算的链路分析基于OpenTracing协议的Jaeger实现，Java运行时提供以下两种创建自定义Span的方式。

使用OpenTelemetry（推荐）

在Java语言的代码中，您可以通过OpenTelemetry SDK手动埋点将数据上报到链路追踪服务端。完整的示例代码，请参见java-tracing-openTelemetry。

示例代码解析如下。

在pom.xml文件中添加依赖。

  <dependencies>
    <dependency>
      <groupId>junit</groupId>
      <artifactId>junit</artifactId>
      <version>3.8.1</version>
      <scope>test</scope>
    </dependency>
    <dependency>
      <groupId>com.aliyun.fc.runtime</groupId>
      <artifactId>fc-java-core</artifactId>
      <version>1.4.1</version>
    </dependency>
    <dependency>
      <groupId>io.opentelemetry</groupId>
      <artifactId>opentelemetry-api</artifactId>
      <version>1.19.0</version>
    </dependency>
    <dependency>
      <groupId>io.opentelemetry</groupId>
      <artifactId>opentelemetry-sdk</artifactId>
      <version>1.19.0</version>
    </dependency>
    <dependency>
      <groupId>io.opentelemetry</groupId>
      <artifactId>opentelemetry-semconv</artifactId>
      <version>1.19.0-alpha</version>
    </dependency>
    <dependency>
      <groupId>io.opentelemetry</groupId>
      <artifactId>opentelemetry-exporter-jaeger-thrift</artifactId>
      <version>1.19.0</version>
    </dependency>
    <dependency>
      <groupId>io.jaegertracing</groupId>
      <artifactId>jaeger-thrift</artifactId>
      <version>1.8.1</version>
    </dependency>
    <dependency>
      <groupId>org.slf4j</groupId>
      <artifactId>slf4j-simple</artifactId>
      <version>1.6.6</version>
    </dependency>
  </dependencies>

上报数据到链路追踪服务端。

public void handleRequest(InputStream inputStream, OutputStream outputStream, Context context) throws IOException {
    String endpoint = context.getTracing().getJaegerEndpoint();
    try {
        ExampleConfiguration.initOpenTelemetry(endpoint);
    } catch (TTransportException e) {
        throw new RuntimeException(e);
    }

    SpanContext spanContext = contextFromString(context.getTracing().getSpanContext());

    startMySpan(io.opentelemetry.context.Context.current().with(Span.wrap(spanContext)));
}

创建一个全局OpenTelemetry对象，提供对Tracers的访问。

static OpenTelemetry initOpenTelemetry(String jaegerEndpoint) throws TTransportException {
    // 导出traces到Jaeger
    JaegerThriftSpanExporter jaegerExporter =
        JaegerThriftSpanExporter.builder()
        .setThriftSender(new Builder(jaegerEndpoint).build())
        .setEndpoint(jaegerEndpoint)
        .build();

    Resource serviceNameResource =
        Resource.create(Attributes.of(ResourceAttributes.SERVICE_NAME, "otel-jaeger-example"));

    // 设置由Jaeger Exporter处理Span
    SdkTracerProvider tracerProvider =
        SdkTracerProvider.builder()
        .addSpanProcessor(SimpleSpanProcessor.create(jaegerExporter))
        .setResource(Resource.getDefault().merge(serviceNameResource))
        .build();
    OpenTelemetrySdk openTelemetry =
        OpenTelemetrySdk.builder().setTracerProvider(tracerProvider).buildAndRegisterGlobal();

    // JVM退出时关闭SDK
    Runtime.getRuntime().addShutdownHook(new Thread(tracerProvider::close));

    return openTelemetry;
}

获取上下文的Tracing信息，转换为SpanContext。

SpanContext contextFromString(String value) throws IOException {
    {
        if (value != null && !value.equals("")) {
            String[] parts = value.split(":");
            if (parts.length != 4) {
                throw new RuntimeException(value);
            } else {
                String traceId = parts[0];
                if (traceId.length() <= 32 && traceId.length() >= 1) {
                    return SpanContext.createFromRemoteParent("0000000000000000"+parts[0], parts[1], TraceFlags.getSampled(), TraceState.getDefault());
                } else {
                    throw new RuntimeException("Trace id [" + traceId + "] length is not withing 1 and 32");
                }
            }
        } else {
            throw new RuntimeException();
        }
    }
}

创建tracer并通过转换的Context创建子Span。每一个Span代表调用链中被命名并计时的连续性执行片段，您也可以基于该Span继续创建子Span。

void startMySpan(io.opentelemetry.context.Context ctx){
    Tracer tracer = GlobalOpenTelemetry.getTracer("fc-Trace");
    Span parentSpan = tracer.spanBuilder("fc-operation").setParent(ctx).startSpan();
    parentSpan.setAttribute("version","fc-v1");
    try {
        TimeUnit.MILLISECONDS.sleep(150);
    } catch (InterruptedException e) {
        throw new RuntimeException(e);
    }
    child(parentSpan.storeInContext(ctx));
    parentSpan.end();
}

void child(io.opentelemetry.context.Context ctx){
    Tracer tracer = GlobalOpenTelemetry.getTracer("fc-Trace");
    Span childSpan = tracer.spanBuilder("fc-operation-child").setParent(ctx).startSpan();
    childSpan.addEvent("timeout");
    try {
        Thread.sleep(100);
    } catch (InterruptedException e) {
        throw new RuntimeException(e);
    }
    childSpan.end();
}

使用Jaeger SDK

您可以通过Jaeger SDK埋点，将数据上报到链路追踪服务端。完整的示例代码，请参见java-tracing。

示例代码解析如下。

在pom.xml文件中添加依赖。

<dependencies>
  <dependency>
    <groupId>junit</groupId>
    <artifactId>junit</artifactId>
    <version>3.8.1</version>
    <scope>test</scope>
  </dependency>
  <dependency>
    <groupId>com.aliyun.fc.runtime</groupId>
    <artifactId>fc-java-core</artifactId>
    <version>1.4.1</version>
  </dependency>
  <dependency>
    <groupId>io.jaegertracing</groupId>
    <artifactId>jaeger-client</artifactId>
    <version>1.8.1</version>
  </dependency>
  <dependency>
    <groupId>org.slf4j</groupId>
    <artifactId>slf4j-simple</artifactId>
    <version>1.6.6</version>
  </dependency>
</dependencies>

上报数据到链路追踪服务端。

public void handleRequest(InputStream inputStream, OutputStream outputStream, Context context) throws IOException {

    registerTracer(context);

    JaegerSpanContext spanContext = contextFromString(context.getTracing().getSpanContext());

    startMySpan(spanContext);
}

根据上下文的Tracing信息创建一个tracer。

void registerTracer(Context context){
    io.jaegertracing.Configuration config = new io.jaegertracing.Configuration("FCTracer");
    io.jaegertracing.Configuration.SenderConfiguration sender = new io.jaegertracing.Configuration.SenderConfiguration();
    sender.withEndpoint(context.getTracing().getJaegerEndpoint());
    config.withSampler(new io.jaegertracing.Configuration.SamplerConfiguration().withType("const").withParam(1));
    config.withReporter(new io.jaegertracing.Configuration.ReporterConfiguration().withSender(sender).withMaxQueueSize(10000));
    GlobalTracer.register(config.getTracer());
}

转换spanContext并创建自定义Span，您也可以基于该Span继续创建子Span。

static JaegerSpanContext contextFromString(String value) throws MalformedTracerStateStringException, EmptyTracerStateStringException {
    if (value != null && !value.equals("")) {
        String[] parts = value.split(":");
        if (parts.length != 4) {
            throw new MalformedTracerStateStringException(value);
        } else {
            String traceId = parts[0];
            if (traceId.length() <= 32 && traceId.length() >= 1) {
                return new JaegerSpanContext(0L, (new BigInteger(traceId, 16)).longValue(), (new BigInteger(parts[1], 16)).longValue(), (new BigInteger(parts[2], 16)).longValue(), (new BigInteger(parts[3], 16)).byteValue());
            } else {
                throw new TraceIdOutOfBoundException("Trace id [" + traceId + "] length is not withing 1 and 32");
            }
        }
    } else {
        throw new EmptyTracerStateStringException();
    }
}