基于Netflix Zuul的代理服务：请求过滤器是否同线程执行？

Great question—this is a common concern when working with MDC and request tracing in Zuul, so let's unpack it clearly for your scenario.

1. Do all Zuul filters for the same request run in the same thread? #

Absolutely, by default all Zuul filters (pre, route, post, error) for a single request execute in the same thread. Zuul uses a synchronous processing model out of the box: when an incoming request hits your proxy, a single thread from the underlying server's pool (like Tomcat's worker threads) handles the entire filter lifecycle for that request.

This is perfect news for your Flow-Id plan! Setting the header into MDC via a pre-filter will make it available across every subsequent filter for the request, since MDC relies on thread-local storage. You won't have to worry about context loss between pre, route, post, or error filters in the default setup.

2. Does this thread handle other tasks during request processing? #

No, during the lifecycle of your specific request, the thread is dedicated solely to running the Zuul filter chain and processing that request until it's completed (response sent back to the client).

That said, once the request finishes, the thread gets returned to the server's thread pool and may be reused for a completely different incoming request later. The only time this thread model changes is if you explicitly implement asynchronous processing in your filters—for example, using ZuulFilter.runAsync() or integrating async handlers. In those cases, parts of the processing might shift to a different thread, which would break MDC continuity unless you manually propagate the context. But if you're sticking to standard synchronous Zuul usage, this isn't something you need to plan for.

Quick Practical Tips for Your Flow-Id Setup #

• Pre-filter Setup: In your pre-filter, extract the Flow-Id header from the incoming request (or generate a unique one if it's missing) and add it to MDC with MDC.put("Flow-Id", flowId).
• Cleanup is Key: Make sure to remove the Flow-Id from MDC in a post-filter or error-filter once the request is done. Use MDC.remove("Flow-Id") to prevent context leakage when the thread is reused for future requests.
• Async Workaround (If Needed): If you ever need to add async logic, you can preserve the Flow-Id by capturing it before switching threads and re-setting it in the async task. Here's a quick example:

  // Capture MDC context before async call
  String flowId = MDC.get("Flow-Id");
  CompletableFuture.runAsync(() -> {
      // Re-set context in the async thread
      MDC.put("Flow-Id", flowId);
      // Your async logic here
      // Cleanup after async work
      MDC.remove("Flow-Id");
  });
  

内容的提问来源于stack exchange，提问作者Thomas