Sync and Async issues

[LongicornYu]

We are using Python 3.6.10 +Gunicorn + FastAPI + SqlAlchemy ORM (similar to this approach https://fastapi.tiangolo.com/tutorial/sql-databases/)

For sync APIs, we experienced a growing memory issue as stated here #596. the memory keeps growing till the pod restart in OpenShift. (One solution is to upgrade to Python 3.8 but we need to stick with Python 3.6 for now)

If we switch all API endpoints to be under async, the growing memory issue is gone, but we experienced some issue that might be related to the comment here: #260 (comment)
As we are using SQLalchemy ORM, it doesn't support async, so essentially we are still calling everything synchronized under the async API endpoints. According to the above-linked comment, some requests might block the main thread and causing other APIs to fail.

Is switching to use SQLalchemy core to work with async APIs the only solution here?

[LongicornYu]

@dmontagu can you collaborate more on the comment you made in this thread #260 (comment)

Appreciate it!

[SnkSynthesis]

You can run your sync functions (like SQLAlchemy) in async contexts like this: (For more info please see this issue)

from fastapi.concurrency import run_in_threadpool

def sync_function():
    ...  # Logic here

async def async_function():
    await run_in_threadpool(sync_function)

Also,
SQLAlchemy will have async support soon.

Though you can also use GINO which is built on top of SQLAlchemy to be async but currently only supports PostgreSQL with the asyncpg driver.

[tiangolo]

Hey there! I created Asyncer to simplify interacting between sync and async code, that might be helpful to you: https://asyncer.tiangolo.com/

Sorry for the long delay! 🙈 I wanted to personally address each issue/PR and they piled up through time, but now I'm checking each one in order.

[github-actions]

Assuming the original need was handled, this will be automatically closed now. But feel free to add more comments or create new issues or PRs.