[Python] Setting lifetime for gunicorn workers

We have a project on FastAPI. It is still very lightly loaded, so after a few hours of inactivity we get an error of this kind . All sorts of recipes with pool_recycle and adding manipulations to the decorator of DB operations do not help. I decided to go around the problem - to turn off workers every hour. I sketched out the following scheme:

main.py

import asyncio
from fastapi import FastAPI
from .workers import worker_shutdown

app = FastAPI()

@app.on_event("startup")
async def worker_lifetime():
loop = asyncio.get_event_loop()
loop.create_task(worker_shutdown())


workers.py

import os
import signal
import asyncio
import random

async def worker_shutdown():
pid = os.getpid()
delay = 3600 + random.randint(2,60)
await asyncio.sleep(delay)
os.kill(pid, signal.SIGTERM)


All this works as intended. The only question is, is it correct? Will it cause any unexpected problems? There was also an idea to move worker_shutdown to a separate thread using threading.

Continue reading...