After a while, the server dies ECONNREFUSED

First Check

• I added a very descriptive title to this issue.
• I used the GitHub search to find a similar issue and didn’t find it.
• I searched the FastAPI documentation, with the integrated search.
• I already searched in Google “How to X in FastAPI” and didn’t find any information.
• I already read and followed all the tutorial in the docs and didn’t find an answer.
• I already checked if it is not related to FastAPI but to Pydantic.
• I already checked if it is not related to FastAPI but to Swagger UI.
• I already checked if it is not related to FastAPI but to ReDoc.

Commit to Help

• I commit to help with one of those options 👆

Example Code

None

Description

I launch fastapi with this command: uvicorn main:app --host 0.0.0.0 --port 17010 --workers 16, and it contains a bunch of endpoints, declared as sync which are passed to Celery and immediately returns, and then Celery does the heavy load.

It has received so far for example approx 100 requests per second, every request looks all alike (same task different parameters), and it works very well and fast. The machine resource consumption is low and everything seems going fine. However, after a few hours, suddenly fastapi stops responding forever: every call is immediately rejected with ECONNREFUSED like it isn’t even listening to the port anymore.

Facts:

• I can see in the taskmanager that the uvicorn process is still running
• I can see in the taskmanager that the 16 processes spawned are still running
• In the uvicorn log, the last line is the last successful request, then nothing
• In the fastapi log, the last line is the last successful request with some other info from celery, then nothing
• The machine has still low resources consumption when it happens
• If I kill all the 17 processes and restart uvicorn, it works as usual
• Always after a few hours it dies again
• There are no weird events in the Windows Event Viewer

I have no clue about what is happening and why I can’t see any crash in the logs. How could I find out more?

Operating System

Windows

Operating System Details

Windows Server 2019, 16 cores, 64GB ram

FastAPI Version

0.63.0

Python Version

3.8.8

Additional Context

No response

[researcher2]

We’ve had it running stable on gunicorn since a bit after my last comment, looks like you simply can’t use uvicorn directly for production.

On Sat, Sep 11, 2021 at 1:29 AM STeve Shary @.***> wrote:

@gygabyte017 https://github.com/gygabyte017 Sorry for the slow response. By large file inputs, I mean uploading a large file. I’m not sure what is the root cause of the problem. I suspect that it is not a FastAPI issue, but rather something with Uvicorn. The way that the connections will work is that they will go from your machine to the server. Then the server kernel will route to Uvicorn. Then Uvicorn will route to a FastAPI instance.

The fact that you get a ECONNREFUSED indicates that the server will not setup a TCP connection to even begin to talk to Uvicorn/FastAPI. It sounds like @researcher2 https://github.com/researcher2 does not face that issue when using nginx which further makes me think that uvicorn has something going on.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/tiangolo/fastapi/issues/3700#issuecomment-916992497, or unsubscribe https://github.com/notifications/unsubscribe-auth/AHULRU24XMY4JBQUJEL42I3UBIP6TANCNFSM5CDW52OA . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.

[github-actions[bot]]

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.