command.stamp hangs when using with uvcorn

[rajshrestha86]

Describe the bug
I am trying to stamp a revision from code in FastAPI. But when I am running command.stamp it hangs there and it doesn't start the API server.

Expected behavior
The alembic should have stamp the db and have started the API server like below:

uvicorn app.main:app --port 8000
INFO:     Started server process [96686]
INFO:     Waiting for application startup.
INFO:     Application startup complete.
INFO:     Uvicorn running on http://127.0.0.1:8000 (Press CTRL+C to quit)

To Reproduce
Here is a snippet of the code that I am working on:

import logging

from alembic import command, config
from fastapi import FastAPI
from sqlmodel import create_engine

DATABASE_URI = "sqlite://"

logging.basicConfig(level=logging.INFO)
log = logging.getLogger(__name__)

app = FastAPI()

sqlmodel_engine = create_engine(DATABASE_URI)


with sqlmodel_engine.begin() as conn:
    alembic_cfg = config.Config("alembic.ini")
    alembic_cfg.attributes["connection"] = conn
    command.stamp(alembic_cfg, "head")

The app hangs with following output when running with uvicorn: uvicorn app.main:app --port 8000

Output

uvicorn app.main:app --port 8000
INFO  [alembic.runtime.migration] Context impl SQLiteImpl.
INFO  [alembic.runtime.migration] Will assume non-transactional DDL.
# Stuck after this

If I comment the line with command.stamp the API server starts successfully as in the first code snippet. If I run the code without uvicorn py just running python main.py it completes it's execution but when I start it with uvcorn it gets stuck. Can't seem to figure out what's the issue here.

[zzzeek]

what if you run a simple query instead of command.stamp, like connection.execute(text("select 1")), does that return?

also:

1. what is the actual database URL and driver in use? You seem to have a default of "sqlite://" memory DB, is that the actual backend you are running on ?
2. what SQLAlchemy version?

[rajshrestha86]

What's interesting is that if I use Flask instead of FastAPI then it works, doesn't hang. But not sure why it hangs when using with FastAPI and uvicorn.

[zzzeek]

OK....using "faulthandler" alone is not enough, as mentioned, when it hangs, keep it hung, then go to your console and do a signal using kill -SIGUSER1 <pid> - that's what triggers faulthandler to print stack where it's hung

[bavarian-ng]

In our case, we had the same problem... I removed the logging options from the alembic.ini as well as the call for loading it from env.py as it seemed to interfere with each other if there were different log levels set in our uvicorn application and alembic.
Maybe this helps someone with same error in the future. :)

[dunielpls]

I came across this issue too, thank you @bavarian-ng. Have anyone managed to figure out why logging.config.fileConfig(config.config_file_name) causes this behaviour?

[dunielpls]

This is caused by fileConfig's argument disable_existing_loggers defaulting to a value of True. Changing the relevant line in env.py to fileConfig(config.config_file_name, disable_existing_loggers=False) solved my issue.