sleuth和zipkin调试流程图（跟踪异步上传）

org.springframework.cloud.sleuth.zipkin.HttpZipkinSpanReporter.Flusher

跟踪进入线程类的run方法

public void flush() {  if (this.pending.isEmpty())   return;  List<Span> drained = new ArrayList<>(this.pending.size());  this.pending.drainTo(drained);  if (drained.isEmpty())   return;  // json-encode the spans for transport  byte[] json = Codec.JSON.writeSpans(drained);  // NOTE: https://github.com/openzipkin/zipkin-java/issues/66 will throw instead of return null.  if (json == null) {   log.debug(\

初始化静态类-定时器1，初始化核心线程数为1的线程池2，spring工程启动时通过spi注入HttpZipkinSpanReporter初始化3，此线程一直异步批量抓取LinkedBlockedQueue中的元素，批量上报zipkin

java.io.Flushable#flush

此类源码如下

frush方法源码如下Calling this will flush any pending spans to the http transport on the current thread

上报调试流程图（跟踪日志异步上传）

跟踪进去postSpans方法，此方法调用zipkin的http接口