spring-cloud Sleuth

摘要： 一直沒弄明白sleuth的tracerContext是如何建立和傳遞的，閒來無事研究了一下。由於對sleuth的原始碼不熟悉，準備通過debug brave.Tracer 的nextId()方法，檢視方法呼叫棧來找來龍去脈。 首先建立兩個service A和B，記作srvA、srvB，...

一直沒弄明白sleuth的tracerContext是如何建立和傳遞的，閒來無事研究了一下。由於對sleuth的原始碼不熟悉，準備通過debug brave.Tracer 的nextId()方法，檢視方法呼叫棧來找來龍去脈。

首先建立兩個service A和B，記作srvA、srvB，在srvA中新增testA controller，sevB中新增testB controller，testA中通過Feign呼叫testB。

1. 先看當用戶通過瀏覽器呼叫srvA的時候，srvA是作為server的。

configuration :

TraceWebServletAutoConfiguration==> TracingFilter

TraceHttpAutoConfiguration==> HttpTracing

TraceAutoConfiguration==> Tracing

SleuthLogAutoConfiguration.Slf4jConfiguration==> CurrentTraceContext

配置中， TracingFilter 在例項化時需要一個 HttpTracing ：

public static Filter create(HttpTracing httpTracing) {
return new TracingFilter(httpTracing);
}

//Servlet執行時類
final ServletRuntime servlet = ServletRuntime.get();
//Slf4jCurrentTraceContext
final CurrentTraceContext currentTraceContext;
final Tracer tracer;
final HttpServerHandler<HttpServletRequest, HttpServletResponse> handler;
//TraceContext的資料提取器
final TraceContext.Extractor<HttpServletRequest> extractor;

TracingFilter(HttpTracing httpTracing) {
tracer = httpTracing.tracing().tracer();
currentTraceContext = httpTracing.tracing().currentTraceContext();
handler = HttpServerHandler.create(httpTracing, ADAPTER);
extractor = httpTracing.tracing().propagation().extractor(GETTER);
}

HttpTracingBuilder模式構造時接收一個 Tracing ：

Tracing tracing;
//客戶端span解析器
HttpClientParser clientParser;
String serverName;
//服務端span解析器
HttpServerParser serverParser;
HttpSampler clientSampler, serverSampler;

Builder(Tracing tracing) {
if (tracing == null) throw new NullPointerException("tracing == null");
final ErrorParser errorParser = tracing.errorParser();
this.tracing = tracing;
this.serverName = "";
// override to re-use any custom error parser from the tracing component
this.clientParser = new HttpClientParser() {
@Override protected ErrorParser errorParser() {
return errorParser;
}
};
this.serverParser = new HttpServerParser() {
@Override protected ErrorParser errorParser() {
return errorParser;
}
};
this.clientSampler = HttpSampler.TRACE_ID;
this.serverSampler(HttpSampler.TRACE_ID);
}

Tracing例項化：

@Bean
@ConditionalOnMissingBean
// NOTE: stable bean name as might be used outside sleuth
Tracing tracing(@Value("${spring.zipkin.service.name:${spring.application.name:default}}") String serviceName,
Propagation.Factory factory,
CurrentTraceContext currentTraceContext,
Reporter<zipkin2.Span> reporter,
Sampler sampler,
ErrorParser errorParser,
SleuthProperties sleuthProperties
) {
return Tracing.newBuilder()
.sampler(sampler)
.errorParser(errorParser)
.localServiceName(serviceName)
//ExtraFieldPropagation.Factory
.propagationFactory(factory)
.currentTraceContext(currentTraceContext)
.spanReporter(adjustedReporter(reporter))
.traceId128Bit(sleuthProperties.isTraceId128())
.supportsJoin(sleuthProperties.isSupportsJoin())
.build();
}

下面看TracingFilter的doFilter：

@Override
public void doFilter(ServletRequest request, ServletResponse response, FilterChain chain)
throws IOException, ServletException {
HttpServletRequest httpRequest = (HttpServletRequest) request;
HttpServletResponse httpResponse = servlet.httpResponse(response);

// Prevent duplicate spans for the same request
TraceContext context = (TraceContext) request.getAttribute(TraceContext.class.getName());
if (context != null) {
// A forwarded request might end up on another thread, so make sure it is scoped
Scope scope = currentTraceContext.maybeScope(context);
try {
chain.doFilter(request, response);
} finally {
scope.close();
}
return;
}

Span span = handler.handleReceive(extractor, httpRequest);

// Add attributes for explicit access to customization or span context
request.setAttribute(SpanCustomizer.class.getName(), span.customizer());
request.setAttribute(TraceContext.class.getName(), span.context());

Throwable error = null;
Scope scope = currentTraceContext.newScope(span.context());
try {
// any downstream code can see Tracer.currentSpan() or use Tracer.currentSpanCustomizer()
chain.doFilter(httpRequest, httpResponse);
} catch (IOException | ServletException | RuntimeException | Error e) {
error = e;
throw e;
} finally {
scope.close();
if (servlet.isAsync(httpRequest)) { // we don't have the actual response, handle later
servlet.handleAsync(handler, httpRequest, httpResponse, span);
} else { // we have a synchronous response, so we can finish the span
handler.handleSend(ADAPTER.adaptResponse(httpRequest, httpResponse), error, span);
}
}
}

在SleuthLogAutoConfiguration中如果有slfj的包，則注入 CurrentTraceContext ：

@Configuration
@ConditionalOnClass(MDC.class)
@EnableConfigurationProperties(SleuthSlf4jProperties.class)
protected static class Slf4jConfiguration {

@Bean
@ConditionalOnProperty(value = "spring.sleuth.log.slf4j.enabled", matchIfMissing = true)
@ConditionalOnMissingBean
public CurrentTraceContext slf4jSpanLogger() {
return Slf4jCurrentTraceContext.create();
}

...
}

Slf4jCurrentTraceContext中，delegate就是 CurrentTraceContext.Default.inheritable() ：

public static final class Default extends CurrentTraceContext {
static final ThreadLocal<TraceContext> DEFAULT = new ThreadLocal<>();
// Inheritable as Brave 3's ThreadLocalServerClientAndLocalSpanState was inheritable
static final InheritableThreadLocal<TraceContext> INHERITABLE = new InheritableThreadLocal<>();

final ThreadLocal<TraceContext> local;

//靜態方法create，local物件為ThreadLocal型別 
/** Uses a non-inheritable static thread local */
public static CurrentTraceContext create() {
return new Default(DEFAULT);
}

//local物件為InheritableThreadLocal型別
//官方文件指出，inheritable方法線上程池的環境中需謹慎使用，可能會取出錯誤的TraceContext，這樣會導致Span等資訊會記錄並關聯到錯誤的traceId上
/**
* Uses an inheritable static thread local which allows arbitrary calls to {@link
* Thread#start()} to automatically inherit this context. This feature is available as it is was
* the default in Brave 3, because some users couldn't control threads in their applications.
*
* <p>This can be a problem in scenarios such as thread pool expansion, leading to data being
* recorded in the wrong span, or spans with the wrong parent. If you are impacted by this,
* switch to {@link #create()}.
*/
public static CurrentTraceContext inheritable() {
return new Default(INHERITABLE);
}

Default(ThreadLocal<TraceContext> local) {
if (local == null) throw new NullPointerException("local == null");
this.local = local;
}

@Override public TraceContext get() {
return local.get();
}

//替換當前TraceContext，close方法將之前的TraceContext設定回去
//Scope介面繼承了Closeable介面，在try中使用會自動呼叫close方法，為了避免使用者忘記close方法，還提供了Runnable，Callable，Executor，ExecutorService包裝方法
@Override public Scope newScope(@Nullable TraceContext currentSpan) {
final TraceContext previous = local.get();
local.set(currentSpan);
class DefaultCurrentTraceContextScope implements Scope {
@Override public void close() {
local.set(previous);
}
}
return new DefaultCurrentTraceContextScope();
}
}

Slf4jCurrentTraceContext的delegate使用的就是一個 InheritableThreadLocal ， InheritableThreadLocal 在建立子執行緒的時候，會將父執行緒的inheritableThreadLocals繼承下來。這樣就實現了 TraceContext 在父子執行緒中的傳遞。

看一下 CurrentTraceContext 的 maybeScope ：

//返回一個新的scope，如果當前scope就是傳入的scope，返回一個空scope
public Scope maybeScope(@Nullable TraceContext currentSpan) {
//獲取當前TraceContext
TraceContext currentScope = get();
//如果傳入的TraceContext為空，且當前TraceContext為空返回空scope
if (currentSpan == null) {
if (currentScope == null) return Scope.NOOP;
return newScope(null);
}
return currentSpan.equals(currentScope) ? Scope.NOOP : newScope(currentSpan);
}

TracingFilter中 HttpServerHandler 解析Request：

請輸入程式碼

2.srvA請求到servB時作為Client。

TraceLoadBalancerFeignClient--> LoadBalancerFeignClient --> FeignLoadBalancer --> LazyTracingFeignClient --> Client