Linux 简短的简单命令如何向进程发送 SIGTERM 并在 X 秒内不存在时发送 SIGKILL？

发布于 2025-01-09 19:58:46 字数 199 浏览 3 评论 0原文

应该如何查看 Linux 命令向进程/PID 发送终止信号，如果它在 10 秒后未能正常退出，则将其杀死？

我的尝试是：“sudo timeout -vk 5 10 Kill PIDhere”（-v verbose，-k Kill after X秒），但我不确定它是否好或如何调整值，或者是否有更好的命令可以使用进程命令行中显示的名称的一部分。（“ps aux”输出）

原文

分享到QQ

分享到微博

如果你对这篇内容有疑问，欢迎到本站社区发帖提问参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。

发布评论

需要登录才能够评论，你可以免费注册一个本站的账号。

够运 2025-01-16 19:58:46

sudo timeout -vk 5 10 kill PIDhere

将执行kill，然后尝试终止该进程（如果需要太长时间）。这不应该发生，并且可能不是您想要的（如果 kill 实际上挂起，杀死它不会影响您的实际进程）。 timeout 对于限制进程运行的时间很有用，而不是收到信号后需要多长时间才能终止。

相反，我建议异步启动进程（例如在 shell 中使用 &，但任何语言的子进程库都将具有类似的功能），然后在向进程发送信号后等待进程终止。我在这个答案中描述了用Java执行此操作。在 shell 中可能如下所示：

$ some_process &
# time passes, eventually we decide to terminate the process
$ kill %1
$ sleep 5s
$ kill -s SIGKILL %1 # will fail and do nothing if %1 has already finished

或者您可以依赖 wait，如果作业在 sleep 完成之前终止，它将提前返回：

$ some_process &
# time passes
$ kill %1
$ sleep 5s &
$ wait -n %1 %2       # returns once %1 or %2 (sleep) complete
$ kill -s SIGKILL %1  # if %2 completes first %1 is still running and will be killed

您可以使用 PID 执行与上面相同的操作它不是作业 ID，只是稍微复杂一点，因为您必须担心 PID 重用。

是否有更好的命令甚至可以使用部分名称

pkill 做你想做的事？

sudo timeout -vk 5 10 kill PIDhere

Will execute kill, and then attempt to terminate that process if it takes too long. Which shouldn't happen, and presumably isn't what you want (if kill was actually hanging, killing it would not affect your actual process). timeout is useful for capping how long a process runs for, not how long it takes to terminate after receiving a signal.

Instead, I'd suggest starting the process asynchronously (e.g. using & in a shell, but any language's subprocess library will have similar functionality) and then waiting for the process to terminate after you send it a signal. I describe doing this in Java in this answer. In the shell that might look like:

$ some_process &
# time passes, eventually we decide to terminate the process
$ kill %1
$ sleep 5s
$ kill -s SIGKILL %1 # will fail and do nothing if %1 has already finished

Or you could rely on wait which will return early if the job terminates before the sleep completes:

$ some_process &
# time passes
$ kill %1
$ sleep 5s &
$ wait -n %1 %2       # returns once %1 or %2 (sleep) complete
$ kill -s SIGKILL %1  # if %2 completes first %1 is still running and will be killed

You can do the same as above with PIDs instead of job IDs, it's just a little more fiddly because you have to worry about PID reuse.