_rust_notify.WatchfilesRustInternalError: error in underlying watcher: IO error for operation on <path>: No such file or directory (os error 2)

[pavdwest]

This error occurs on startup in a docker environment (based on python:3.10-slim) when using the --reload flag.

The path is pointing to the python3 executable of my virtual environment i.e. /backend/app/.ignore/venv/bin/python3.

I've tried using the --reload-dir /backend/app/src/ option to entirely exclude the .../.ignore/... directory but it doesn't help - seems it's specifically trying to look for the python3 path?

I've been using this setup for a long time without issue - the only thing that I'm sure has changed is I've upgraded to the latest version of Uvicorn (0.30.1). I've tried resolving this by downgrading as far as 0.25.0 but that doesn't resolve the issue. So I'm very confused now. Any ideas would be welcomed.

Full stack trace:

Attaching to backend-1
backend-1  | INFO:     Will watch for changes in these directories: ['/backend/app/src']
backend-1  | INFO:     Uvicorn running on http://0.0.0.0:8000 (Press CTRL+C to quit)
backend-1  | INFO:     Started reloader process [1] using WatchFiles
backend-1  | Traceback (most recent call last):
backend-1  |   File "/opt/venv/bin/uvicorn", line 8, in <module>
backend-1  |     sys.exit(main())
backend-1  |   File "/opt/venv/lib/python3.10/site-packages/click/core.py", line 1157, in __call__
backend-1  |     return self.main(*args, **kwargs)
backend-1  |   File "/opt/venv/lib/python3.10/site-packages/click/core.py", line 1078, in main
backend-1  |     rv = self.invoke(ctx)
backend-1  |   File "/opt/venv/lib/python3.10/site-packages/click/core.py", line 1434, in invoke
backend-1  |     return ctx.invoke(self.callback, **ctx.params)
backend-1  |   File "/opt/venv/lib/python3.10/site-packages/click/core.py", line 783, in invoke
backend-1  |     return __callback(*args, **kwargs)
backend-1  |   File "/opt/venv/lib/python3.10/site-packages/uvicorn/main.py", line 410, in main
backend-1  |     run(
backend-1  |   File "/opt/venv/lib/python3.10/site-packages/uvicorn/main.py", line 572, in run
backend-1  |     ChangeReload(config, target=server.run, sockets=[sock]).run()
backend-1  |   File "/opt/venv/lib/python3.10/site-packages/uvicorn/supervisors/basereload.py", line 52, in run
backend-1  |     for changes in self:
backend-1  |   File "/opt/venv/lib/python3.10/site-packages/uvicorn/supervisors/basereload.py", line 71, in __next__
backend-1  |     return self.should_restart()
backend-1  |   File "/opt/venv/lib/python3.10/site-packages/uvicorn/supervisors/watchfilesreload.py", line 84, in should_restart
backend-1  |     changes = next(self.watcher)
backend-1  |   File "/opt/venv/lib/python3.10/site-packages/watchfiles/main.py", line 121, in watch
backend-1  |     raw_changes = watcher.watch(debounce, step, rust_timeout, stop_event)
backend-1  | _rust_notify.WatchfilesRustInternalError: error in underlying watcher: IO error for operation on /backend/app/.ignore/venv/bin/python3: No such file or directory (os error 2)

[pavdwest]

Found the issue: when adding python-dateutil as a dependency this error happens. Now I'm even more confused.

[pavdwest]

And the issue is back when upgrading to python 3.11

[lourestc]

Downgrading watchfiles to version 0.21.0 solved the issue for me.

[Kludex]

Please create an issue in watchfiles repository then 🙏

[pavdwest]

Great - thanks @lourestc. I didn't have watchfiles installed as a specific dependency but specifically adding that (version 0.21.0) to my project has resolved the issue.

Also created issue in watchfiles: samuelcolvin/watchfiles#282