Mac M2 Air上Java线程池最大线程数限制问题排查

我正在为异步线程缓存功能做POC验证，目标是在Mac M2 Air上将线程均匀分配到多个核心。但无论分配多少JVM内存，创建线程的数量都存在上限，想了解这是Mac OS的问题还是我遗漏了什么？

参考程序：

/**
This is a simple Java program to demonstrate thread pooling with fixed number of threads using the Executor framework. The program creates a fixed thread pool with a specified number of threads, assigns each thread to a processor, and prints the number of threads running on each processor. */

import java.util.HashMap;
import java.util.Map;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;

public class Main {
    public static void main(String[] args) { 
        // Define number of threads 
        int numThreads = 2000; 
        // Create a fixed thread pool with specified number of threads 
        ExecutorService executor = Executors.newFixedThreadPool(numThreads);

        // Set the number of processors to be used by the JVM
        int numProcessors = 4; // Arbitrary value  
        System.setProperty("java.util.concurrent.ForkJoinPool.common.parallelism", Integer.toString(numProcessors));

        // Create a HashMap to count threads assigned to each processor
        Map<Integer, Integer> processorThreadCount = new HashMap<>();
        for (int i = 0; i < numProcessors; i++) {
            processorThreadCount.put(i, 0);
        }

        // Assign each thread to a processor and execute it
        for (int i = 0; i < numThreads; i++) {
            int processor = i % numProcessors;
            processorThreadCount.compute(processor, (k, v) -> v + 1);
            executor.execute(new MyThread(i));
        }

        // Shutdown the thread pool
        executor.shutdown();

        // Print the number of threads running on each processor
        System.out.println("Threads per processor:");
        processorThreadCount.forEach((k, v) -> System.out.println("Processor " + k + ": " + v));

    }

    /**
    This is a simple thread class that prints the thread ID and the processor it is running on. */

    private static class MyThread implements Runnable {

        private int id;

        public MyThread(int id) { 
            this.id = id;
        }

        public void run() {
            System.out.println("Thread " + id + " running on processor " + Thread.currentThread().getName()); 
        }
    }
}

当将线程数修改为9000时：

int numThreads = 9000;

出现错误：

[0.167s][warning][os,thread] Failed to start thread "Unknown thread" - pthread_create failed (EAGAIN) for attributes:
 stacksize: 102400k, guardsize: 16k, detached.

[0.167s]
[warning][os,thread] Failed to start the native thread for java.lang.Thread "pool-1-thread-2022"Exception in thread "main" 

java.lang.OutOfMemoryError: unable to create native thread: possibly out of memory or process/resource limits reachedat 

java.base/java.lang.Thread.start0(Native Method) at 

java.base/java.lang.Thread.start(Thread.java:803) at 

java.base/java.util.concurrent.ThreadPoolExecutor.addWorker(ThreadPoolExecutor.java:937) at 

java.base/java.util.concurrent.ThreadPoolExecutor.execute(ThreadPoolExecutor.java:1343)at Main.main(Main.java:151)

^C%

我尝试通过java -Xss10m Main将JVM栈大小从初始的1Mb调整到10Mb，甚至增加到1GB，但每次都会出现pool-1-thread-2022线程分配失败的OutOfMemoryError错误。想知道是什么导致了这个限制？

当线程数设为2000时，程序输出如下：

Threads per processor:
Processor 0: 500
Processor 1: 500
Processor 2: 500
Processor 3: 500

1. MacOS的进程线程数系统限制 #

MacOS对单个进程能创建的原生线程数量有硬性限制，这个限制和JVM堆内存无关。你可以通过以下命令查看当前限制：

• 查看用户级线程数上限：ulimit -u
• 查看系统级线程数上限：sysctl kern.num_taskthreads

当进程创建的线程数接近这个上限时，系统会拒绝创建新线程，抛出pthread_create failed (EAGAIN)错误，也就是你看到的异常。

2. 线程栈内存的反向影响 #

每个Java线程对应一个系统原生线程，每个线程都有独立的栈空间。你调整-Xss增大栈大小，会让每个线程占用更多虚拟内存，反而更快耗尽进程的内存配额，导致无法创建更多线程。正确的做法是调小栈大小（比如-Xss256k），这样每个线程占用的内存更少，能创建的线程数会更多。

3. 线程池设计的误区 #

Executors.newFixedThreadPool(numThreads)会一次性创建所有指定数量的线程，这在线程数很大时是非常低效的——M2 Air是8核（4性能核+4能效核），几千个线程会导致频繁的上下文切换，反而严重降低性能。

实际场景中，线程池大小应该根据任务类型设置：

• CPU密集型任务：线程数≈CPU核心数
• IO密集型任务：线程数≈CPU核心数×(1+IO耗时/CPU耗时)

你想让线程均匀分配到核心，完全不需要创建几千个线程，少量线程就能占满所有核心的计算资源。

4. 代码中处理器分配逻辑无效 #

你用i % numProcessors只是做了计数统计，并没有真正将线程绑定到指定处理器。线程的调度完全由操作系统负责，ForkJoinPool.common.parallelism的设置只对ForkJoinPool生效，对FixedThreadPool没有作用。

如果确实需要将线程绑定到特定核心，Java标准库没有直接支持，需要借助JNI或非标准API（比如OpenJDK的jdk.internal.misc.Unsafe），但这种操作不推荐在生产环境使用，因为会破坏JVM的跨平台性，且对性能提升有限。

内容的提问来源于stack exchange，提问作者ipmob