在MacOS上用nox运行Python测试时出现ZombieProcess异常的原因

问题场景 #

本地通过python3 -m nox运行Python测试，测试用例全部通过，但每次会话都会抛出psutil.ZombieProcess异常；相同配置在GitHub Actions上运行完全正常，无任何相关错误提示。异常触发点在自定义的进程清理逻辑中，调用proc.name()时抛出。

异常栈信息 #

======================================================================== 1 passed in 0.01s ========================================================================
nox > Session my_boto3_test(boto3_version='1.22.13') raised exception psutil.ZombieProcess(pid=86804, msg="PID still exists but it's a zombie")
Traceback (most recent call last):
  File "/project_path/venv/lib/python3.9/site-packages/psutil/_psosx.py", line 343, in wrapper
    return fun(self, *args, **kwargs)
  File "/project_path/venv/lib/python3.9/site-packages/psutil/_psosx.py", line 401, in cmdline
    return cext.proc_cmdline(self.pid)
ProcessLookupError: [Errno 3] assume no such process (originated from sysctl(KERN_PROCARGS2) -> EINVAL)

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/project_path/venv/lib/python3.9/site-packages/nox/sessions.py", line 695, in execute
    self.func(session)
  File "/project_path/venv/lib/python3.9/site-packages/nox/_decorators.py", line 113, in __call__
    return super().__call__(*args, **kwargs)
  File "/project_path/venv/lib/python3.9/site-packages/nox/_decorators.py", line 68, in __call__
    return self.func(*args, **kwargs)
  File "/project_path/venv/lib/python3.9/site-packages/nox/_decorators.py", line 68, in __call__
    return self.func(*args, **kwargs)
  File "/project_path/noxfile.py", line 191, in integration_tests_boto3_sqs
    kill_process_and_clean_outputs(full_path, "run_app", session)
  File "/project_path/noxfile.py", line 647, in kill_process_and_clean_outputs
    kill_process(process_name)
  File "/project_path/noxfile.py", line 657, in kill_process
    if proc.name() == process_name:
  File "/project_path/venv/lib/python3.9/site-packages/psutil/__init__.py", line 628, in name
    cmdline = self.cmdline()
  File "/project_path/venv/lib/python3.9/site-packages/psutil/__init__.py", line 681, in cmdline
    return self._proc.cmdline()
  File "/project_path/venv/lib/python3.9/site-packages/psutil/_psosx.py", line 346, in wrapper
    raise ZombieProcess(self.pid, self._name, self._ppid)
psutil.ZombieProcess: PID still exists but it's a zombie (pid=86804)
nox > Session my_boto3_test(boto3_version='1.22.13') failed.

原因分析 #

僵尸进程是已终止但父进程未回收其资源的进程，本地环境与GitHub Actions容器环境的进程回收机制存在差异：GitHub Actions容器中进程退出后会被快速回收，而本地系统可能存在回收延迟，导致psutil遍历进程时碰到僵尸进程并抛出异常。

解决办法 #

1. 捕获ZombieProcess异常，跳过僵尸进程

修改kill_process函数，在访问进程属性时捕获该异常，直接跳过无需处理的僵尸进程：

import psutil
from psutil import ZombieProcess

def kill_process(process_name):
    for proc in psutil.process_iter(['pid', 'name']):
        try:
            if proc.name() == process_name:
                proc.kill()
        except ZombieProcess:
            # 僵尸进程已死亡，无需处理
            continue

2. 优化进程清理逻辑，避免依赖单一属性

结合进程名和命令行参数匹配进程，同时添加进程终止后的等待逻辑，减少僵尸进程产生：

import psutil
from psutil import ZombieProcess, TimeoutExpired, AccessDenied

def kill_process(process_name):
    for proc in psutil.process_iter(['pid', 'name', 'cmdline']):
        try:
            # 多维度匹配进程，降低误杀概率
            if proc.name() == process_name and 'run_app' in ' '.join(proc.cmdline()):
                proc.terminate()
                # 等待进程正常终止，超时则强制杀死
                proc.wait(timeout=3)
        except (ZombieProcess, TimeoutExpired, AccessDenied):
            continue

3. 升级psutil版本

旧版本psutil在处理僵尸进程时可能存在兼容性问题，升级到最新版本可修复部分场景的异常：

pip install --upgrade psutil

4. 调整nox会话的进程回收策略

在nox会话函数末尾添加显式的进程清理逻辑，确保子进程被正确回收：

@nox.session
def my_boto3_test(session):
    # 测试逻辑
    session.run("pytest", "...")
    # 会话结束前清理进程
    kill_process("run_app")

说明 #

该异常属于本地环境进程回收延迟导致的非功能性错误，不影响测试结果，但会导致nox会话被标记为失败。通过上述方法可避免异常抛出，让nox会话正常结束。

内容的提问来源于stack exchange，提问作者therightstuff