ASP.NET Core 2.0站点约50次请求后卡顿问题求助

Hey there, let's tackle this frustrating cyclic slowdown issue you're hitting. It's really strange that even after stripping out all business logic, you're still seeing this pattern—50 requests leading to 40-second delays, then a 5-minute idle period fixing things, only to repeat. Let's break down actionable troubleshooting steps to get to the root cause:

1. Monitor IIS Express and .NET Process Resource Usage #

First, let's rule out resource exhaustion:

• Open Task Manager and keep an eye on iisexpress.exe (and dotnet.exe if you're using OutOfProcess hosting) as you run your 50 requests. Check if memory usage spikes dramatically, or CPU stays pinned at 100% during the slowdown. If memory climbs and doesn't release, even with no business logic, there might be a hidden leak in the base framework or hosting layer.
• Dig into your IIS Express configuration: Navigate to your project's .vs\config\applicationhost.config file, find your site's <application> node, and check the <recycling> settings. Look for memoryLimit (percentage of system memory that triggers a recycle) or idleTimeout—if the process hits the memory limit before recycling, it could cause slowdowns as it struggles to handle requests while preparing to restart.

2. Check Thread Pool and Async Operation Issues #

ASP.NET Core relies heavily on the thread pool, and even small misconfigurations or unhandled async operations can cause bottlenecks:

• Use Visual Studio's Diagnostic Tools (enable it before running the app) to monitor thread pool activity. Watch the "ThreadPool Threads" and "Work Items Queued" metrics as you send requests. If the queue length keeps growing with each request, it means the thread pool can't keep up—this could be from unawaited async methods in your base code, or a framework-level bug in ASP.NET Core 2.0 that's causing thread leaks.
• Double-check your minimal code: Even if you removed business logic, do you have any middleware or startup code that uses async without proper await? For example, a middleware that calls an async method but doesn't await it could leave hanging operations that clog the thread pool over time.

3. Verify Socket/Port Resource Exhaustion #

Frequent rapid requests can lead to exhausted network resources, especially on Windows:

• When the site starts slowing down, open Command Prompt and run netstat -ano | findstr :<your-site-port>. Look for a large number of connections in TIME_WAIT state. Windows recycles these ports automatically, but if you're hitting requests faster than the recycle rate, it can cause new requests to wait for available ports—leading to those 40-second delays.
• You can temporarily adjust Windows' port recycling settings to test if this resolves the issue, though this is more of a workaround than a root fix.

4. Test with Ultra-Minimal Code and Rule Out Framework Bugs #

Since you already stripped out business logic, go even further to isolate the issue:

• Create a brand new ASP.NET Core 2.0 project with the absolute minimum: In Startup.cs, remove all middleware except a simple inline middleware that returns a plain string, like:

  app.Run(async (context) =>
  {
      await context.Response.WriteAsync("Hello World!");
  });
  

• Run this bare-bones app and repeat your 50-request test. If the slowdown still happens, it's likely a bug or compatibility issue with ASP.NET Core 2.0 and your version of IIS Express/Windows. Keep in mind ASP.NET Core 2.0 is end-of-life, so upgrading to a newer supported version (even 3.1 LTS) might resolve the issue entirely.

5. Disable Debugger and Test Release Mode #

Sometimes the Visual Studio debugger can introduce unexpected overhead or resource locks:

• Run your app in Release mode without attaching the debugger (use Ctrl+F5 instead of F5). If the slowdown disappears, the issue might be related to debugger instrumentation rather than your code or hosting setup.

内容的提问来源于stack exchange，提问作者hdimon