Please add scalable app file structure in documentation. Also is my approach correct?

• asyncpg version:
• PostgreSQL version:
• Do you use a PostgreSQL SaaS? If so, which? Can you reproduce
  the issue with a local PostgreSQL install?:
• Python version:
• Platform:
• Do you use pgbouncer?:
• Did you install asyncpg with pip?:
• If you built asyncpg locally, which version of Cython did you use?:
• Can the issue be reproduced under both asyncio and
  uvloop?:

Latest version of asyncpg
No PGBouncer
Python 3.8.2

Hi, I am using fast API but I have no idea how to maintain DB connection and pool. I am switching from Django ORM but still, it is hard for me to get a good grasp on this. Please guide me on how can I improve this and also add this to your documentation.

My questions:
How to maintain a constant database connection?
How to maintain a pool of connections?
How to use the same connection out of the pool for every query in the same request?

Is my naive solution correct? How can I improve this and make this production-ready?

db_connection.py

`import asyncpg
from configs.settings import settings


class Database:
    def __init__(self):
        self.user = settings.POSTGRES_USER
        self.password = settings.POSTGRES_PASSWORD
        self.host = settings.POSTGRES_SERVER
        self.port = "5432"
        self.database = settings.POSTGRES_DB
        self._cursor = None

        self._connection_pool = None
        self.con = None

    async def connect(self):
        if not self._connection_pool:
            try:
                self._connection_pool = await asyncpg.create_pool(
                    min_size=1,
                    max_size=10,
                    command_timeout=60,
                    host=self.host,
                    port=self.port,
                    user=self.user,
                    password=self.password,
                    database=self.database,
                )

            except Exception as e:
                print(e)

    async def fetch_rows(self, query: str):
        print(query)
        if not self._connection_pool:
            print("shouldnt be here")
            await self.connect()
        else:
            self.con = await self._connection_pool.acquire()
            try:
                result = await self.con.fetch(query)
                print(result)
                return result
            except Exception as e:
                print(e)
            finally:
                print("pool released")
                await self._connection_pool.release(self.con)

`

db_session.py

`from .db_connection import Database

database_instance = Database()

`

main.py

`from fastapi import FastAPI
from fastapi.middleware.cors import CORSMiddleware


from routes import items, user

from utils.middleware import middleware
from configs import open_api_tags
from configs.settings import settings
from db.db_session import database_instance

app = FastAPI(
    title=settings.PROJECT_NAME,
    description=settings.PROJECT_DESCRIPTION,
    version="0.0.1",
    openapi_tags=open_api_tags.tags_metadata,
)


app.add_middleware(
    CORSMiddleware,
    allow_origins=settings.BACKEND_CORS_ORIGINS,
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)


app.middleware("http")(middleware)


@app.on_event("startup")
async def startup():
    await database_instance.connect()


app.include_router(user.router)
app.include_router(items.router, prefix="/items")
`

user.py

`from fastapi import APIRouter
from db.db_session import database_instance

router = APIRouter()

@router.get("/users/me", tags=["users"])
async def read_user_me():
    result = await database_instance.fetch_rows("SELECT * from user")
    print(result)
    return {"username": "fakecurrentuser"}`

Start by reviewing the proposed db_connection.py, db_session.py, main.py, and user.py structure alongside the existing asyncpg documentation. Clarify the documented approach for maintaining a pool, acquiring and releasing connections, and using one connection across a request. Done means the documentation gives a production-oriented scalable structure and answers the three connection-lifecycle questions.