subproc_mgr

Motivation

Applications that spawn long-running subprocesses need to be able to manage the lifetime of those subprocesses, which is so-far not a truth as the application can exit unexpectedly (e.g., runtime, or buggy high-level software).

This is a simple service that can be deployed globally that proxies subprocess spawning and manages the lifetime. The service will periodically check the lifetime of the subprocesses and kill them if they have exceeded the expected lifetime.

P.S: This approach could be better (easier and cheaper) than a watchdog if you have a consensus conventions on how to spawn subprocesses.

Usage

1. The service should be started:

   usage: subproc_mgr [-h] [--port PORT] [--period PERIOD] [port] [period]
   
   
       port: port that the service is deployed on
       period: how often the service checks the lifetime of subprocesses
   
   
   positional arguments:
       port             <class 'int'>
       period           <class 'float'>
   
   options:
       -h, --help       show this help message and exit
       --port PORT      <class 'int'>
       --period PERIOD  <class 'float'>
   

2. We send requests (e.g., HTTP requests) to use the service:

   shell> cat usage_example.py
   shell> python usage_example.py
   

   from typing import TypedDict
   import requests, os, shutil
   
   PORT = 5687
   
   class StartProcessRequest(TypedDict):
       command: str
       args: list[str]
       env: dict[str, str]
       pid: int
   
   # start python -m http.server
   resp = requests.post(
       f"http://localhost:{PORT}/spawn",
       json=StartProcessRequest(
           command=shutil.which("python") or "python",
           args=['-m', 'http.server'],
           env=dict(os.environ),
           pid=os.getpid()
       )
   )
   
   assert resp.json().get("code") == "success"
   
   # press Ctrl+D, the subprocess exits very sooner
   while True:
       import time
       time.sleep(0.5)
   

3. Press ctrl+D for the terminal that executes python usage_example.py. You'll see the following logs from the service.

   

Motivation (for the author)

1. Extending VSCode with dynamically added commands/key bindings.
2. Starting a language runtime and managing the lifetime of its parser server.