uvicorn.run() not respecting the reload_excludes parameter?

[abceleung]

I am using FastAPI with Python built-in logging module.

Whenever the log file is updated, the program reloads as well, creating an infinite loop.

(I am using uvicorn.run() for debugging as suggested in the FastAPI doc)

Maybe I made a mistake somewhere? Here is my script:

from fastapi import FastAPI
import uvicorn, logging
from logging.handlers import TimedRotatingFileHandler

logger = logging.getLogger()
logger.setLevel(logging.INFO)
logHandler = TimedRotatingFileHandler(
    "api.log",
    when="D",
    backupCount=10,
    encoding="utf-8",
)
logHandler.setFormatter(
    logging.Formatter(
        "%(asctime)s %(levelname)-8s %(name)-15s %(message)s",
        datefmt="%H:%M:%S",
    )
)
logger.addHandler(logHandler)

app = FastAPI()

@app.get("/")
async def root():
    return {"message": "Hello World"}
    
if __name__ == "__main__":
    uvicorn.run(
        "__main__:app",
        host="0.0.0.0",
        port=8000,
        reload=True,
        reload_excludes="*.log"
    )

Note: I have watchfiles installed.

[euri10]

how do you run uvicorn ? what exclude parameters are you passing ? what is the directory structure like, where the log file is located ?

[abceleung]

how do you run uvicorn ? what exclude parameters are you passing ? what is the directory structure like, where the log file is located ?

how do you run uvicorn ? what exclude parameters are you passing ? what is the directory structure like, where the log file is located ?

1. python main.py <- main.py is the filename
2. reload_excludes="*.log", as shown in the script
3. As shown below (I am using Windows 10):

project
│   main.py
│   api.log (Generated by logger)   
│
└───.venv
      │   python libraries...

[euri10]

cant reproduce on linux though the api.log file is filled wuite rapidely with 10:30:06 INFO watchfiles.main 1 change detected, I seem to recall there was something with watchfiles on windows, maybe someone will remember

[abceleung]

api.log file is filled wuite rapidely with 10:30:06 INFO watchfiles.main 1 change detected

This is the infinite loop I am talking about

[euri10]

reload_excludes just tells uvicorn not to trigger a reload if there is a change in the file, which is what is happening: api.log is changing and uvicorn does not reload itself, so everything works fine to me.

[Bazzu85]

can you please share how to exclude this files from logging config?

[ridham-infocusp]

one way I have achieved infinite logging is by creating a filter and adding that to the root logger

class IgnoreChangeDetectedFilter(logging.Filter):
def filter(self, record: logging.LogRecord):
return '%d change%s detected: %s' != record.msg

root_logger = logging.getLogger()
root_logger.addFilter(IgnoreChangeDetectedFilter())

[obrua]

one way I have achieved infinite logging is by creating a filter and adding that to the root logger

class IgnoreChangeDetectedFilter(logging.Filter): def filter(self, record: logging.LogRecord): return '%d change%s detected: %s' != record.msg

root_logger = logging.getLogger() root_logger.addFilter(IgnoreChangeDetectedFilter())

lookthis:
`
for name in logging.root.manager.loggerDict.keys():
logging.getLogger(name).handlers = []
if "watchfiles.main" in name:
logging.getLogger(name).propagate = False
else:
logging.getLogger(name).propagate = True

#logging.info(f"{logging.getLogger(name)}, {logging.getLogger(name).propagate}")
`

[socrateslee]

The log is from a function in watchfiles

def _log_changes(changes: Set[FileChange]) -> None:
    if logger.isEnabledFor(logging.INFO):  # pragma: no branch
        count = len(changes)
        plural = '' if count == 1 else 's'
        if logger.isEnabledFor(logging.DEBUG):
            logger.debug('%d change%s detected: %s', count, plural, changes)
        else:
            logger.info('%d change%s detected', count, plural)

Setting the logging level for logger watchfiles.main could disable it

import logging
logger = logging.getLogger('watchfiles.main')
logger.setLevel(logging.WARNING)

[vavilon2142]

had the same issue, the problem is that when flag "reload=True" and your logs file in the same directory where app is, it constantly detects new logs and writes the log into file.
So the solution is to allocate logs file outside the app directory
for instance: your directory for app is /app/main.app then logs should be outside the app folder. put them to /opt/logs for example