java - Spring 中的并行请求数与 http-nio-8080-exec 线程数

Question

我想在我的 Controller 中测试并行请求。我知道默认情况下Spring可以处理200个并行请求，我们可以通过修改这个属性server.tomcat.max-threads

来改变它

我稍微研究了一下这个值，发现了有趣的事情:当我启动应用程序时将其设置为 3 时，我看到正在创建 3 个线程:http-nio-8080-exec-1,2,3当我将其设置为 5 时，我会看到 5 个这样的线程。它一直持续到 10，并在 10 处停止。当我将其设置为 15 时，仍然有 10 个名为 http-nio-8080-exec 的线程。有人能解释一下为什么它永远不会超过10吗？

如果我要制作这样的 Controller

@GetMapping("test")
public String test(@RequestParam("skip") boolean skip) throws InterruptedException {
    if(!skip) {
        System.out.println("I'm starting waiting");
        Thread.sleep(10000);
        System.out.println("I stopped waiting");
    }
    return "dd";
}

当有 server.tomcat.max-threads=200

发出 10 个这样的请求:http://localhost:8080/test?skip=false

同时第 11 个请求:http://localhost:8080/test?skip=true

前 10 个请求在返回响应之前应等待 10 秒，但第 11 个请求应立即返回 - 问题是第 11 个请求也在等待，因此被阻止。有人可以解释它是如何工作的吗？如果我将 server.tomcat.max-threads 设置为 200，那么我希望能够处理 200 个独立请求，对吗？

Answer 1

我对这种行为感到好奇并亲自测试了它。不幸的是，我无法确认您正在经历的行为。为了能够同时测试对服务的请求，我编写了一个简单的 go 程序来证明响应时间。因此，我启动了一个带有单个休息端点的 Spring Boot 服务，其中包含您上面发布的代码，并作为客户端遵循 go 程序:

package main

import (
    "log"
    "os/exec"
    "time"
)

func main() {
    for i := 0; i< 15; i++ {
        go runCmd(i)
    }
    time.Sleep(time.Second * 50)
}

func runCmd(i int) {
    param := "localhost:8080/test?skip=false"
    if i > 10 {
        param = "localhost:8080/test?skip=true"
    }
    cmd := exec.Command("curl", param)
    log.Printf("Running command %d and waiting for it to finish...", i)
    start := time.Now()
    err := cmd.Run()
    end := time.Now()
    duration := end.Sub(start)
    log.Printf("Command  %d finished within of second %f error: %v", i, duration.Seconds(), err)
}

该程序使用curl 发送参数skip=false 的前10 个请求，以及skip=true 的另外5 个请求。输出是:

2020/03/08 19:21:18 Running command 14 and waiting for it to finish...
2020/03/08 19:21:18 Running command 9 and waiting for it to finish...
2020/03/08 19:21:18 Running command 3 and waiting for it to finish...
2020/03/08 19:21:18 Running command 10 and waiting for it to finish...
2020/03/08 19:21:18 Running command 7 and waiting for it to finish...
2020/03/08 19:21:18 Running command 0 and waiting for it to finish...
2020/03/08 19:21:18 Running command 12 and waiting for it to finish...
2020/03/08 19:21:18 Running command 6 and waiting for it to finish...
2020/03/08 19:21:18 Running command 5 and waiting for it to finish...
2020/03/08 19:21:18 Running command 13 and waiting for it to finish...
2020/03/08 19:21:18 Running command 11 and waiting for it to finish...
2020/03/08 19:21:18 Running command 2 and waiting for it to finish...
2020/03/08 19:21:18 Running command 1 and waiting for it to finish...
2020/03/08 19:21:18 Running command 4 and waiting for it to finish...
2020/03/08 19:21:18 Running command 8 and waiting for it to finish...
2020/03/08 19:21:18 Command  12 finished within of second 0.035109 error: <nil>
2020/03/08 19:21:18 Command  14 finished within of second 0.035290 error: <nil>
2020/03/08 19:21:18 Command  11 finished within of second 0.040090 error: <nil>
2020/03/08 19:21:18 Command  13 finished within of second 0.041358 error: <nil>
2020/03/08 19:21:28 Command  9 finished within of second 10.034510 error: <nil>
2020/03/08 19:21:28 Command  0 finished within of second 10.034436 error: <nil>
2020/03/08 19:21:28 Command  10 finished within of second 10.037470 error: <nil>
2020/03/08 19:21:28 Command  6 finished within of second 10.042294 error: <nil>
2020/03/08 19:21:28 Command  5 finished within of second 10.042328 error: <nil>
2020/03/08 19:21:28 Command  7 finished within of second 10.045510 error: <nil>
2020/03/08 19:21:28 Command  3 finished within of second 10.045638 error: <nil>
2020/03/08 19:21:28 Command  1 finished within of second 10.049024 error: <nil>
2020/03/08 19:21:28 Command  2 finished within of second 10.053824 error: <nil>
2020/03/08 19:21:28 Command  8 finished within of second 10.053102 error: <nil>
2020/03/08 19:21:28 Command  4 finished within of second 10.053306 error: <nil>

正如您在输出请求 11 - 14 中看到的那样，请求 11 - 14 没有延迟地完成，因此它们没有被阻止。所以可能你必须检查你的测试，可能问题出在测试中，而不是在 tomcat 配置中。

处理 200 个独立请求的期望是正确的，因此它应该表现出来并且确实如此。