[QUESTION] FastAPI keeps booting new workers when deploying

Hi there,

I’m trying to run two different instances of FastAPI in the same Kubernetes pod (different Docker containers), but one of the two seems to keep booting up new workers. Therefor, the container uses a lot of CPU for nothing. It’s also reporting that it’s waiting for application startup constantly, but it is reachable (but slow).

Here is the log that I keep getting:

Checking for script in /app/prestart.sh
Running script /app/prestart.sh
Running inside /app/prestart.sh, you could add migrations to this file, e.g.:
 #! /usr/bin/env bash
 # Let the DB start
sleep 10;
# Run migrations
alembic upgrade head
 [2019-05-24 11:51:24 +0000] [1] [INFO] Starting gunicorn 19.9.0
[2019-05-24 11:51:24 +0000] [1] [INFO] Listening at: http://0.0.0.0:8081 (1)
[2019-05-24 11:51:24 +0000] [1] [INFO] Using worker: uvicorn.workers.UvicornWorker
[2019-05-24 11:51:24 +0000] [8] [INFO] Booting worker with pid: 8
[2019-05-24 11:51:24 +0000] [9] [INFO] Booting worker with pid: 9
WARNING:root:email-validator not installed, email fields will be treated as str.
To install, run: pip install email-validator
WARNING:root:email-validator not installed, email fields will be treated as str.
To install, run: pip install email-validator
[2019-05-24 11:51:33 +0000] [9] [INFO] Started server process [9]
[2019-05-24 11:51:33 +0000] [9] [INFO] Waiting for application startup.
[2019-05-24 11:51:33 +0000] [8] [INFO] Started server process [8]
[2019-05-24 11:51:33 +0000] [8] [INFO] Waiting for application startup.
[2019-05-24 11:51:37 +0000] [18] [INFO] Booting worker with pid: 18
WARNING:root:email-validator not installed, email fields will be treated as str.
To install, run: pip install email-validator
[2019-05-24 11:51:44 +0000] [18] [INFO] Started server process [18]
[2019-05-24 11:51:44 +0000] [18] [INFO] Waiting for application startup.
[2019-05-24 11:51:45 +0000] [23] [INFO] Booting worker with pid: 23
WARNING:root:email-validator not installed, email fields will be treated as str.
To install, run: pip install email-validator
[2019-05-24 11:51:53 +0000] [23] [INFO] Started server process [23]
[2019-05-24 11:51:53 +0000] [23] [INFO] Waiting for application startup.
[2019-05-24 11:51:54 +0000] [28] [INFO] Booting worker with pid: 28
WARNING:root:email-validator not installed, email fields will be treated as str.
To install, run: pip install email-validator
[2019-05-24 11:52:02 +0000] [28] [INFO] Started server process [28]
[2019-05-24 11:52:02 +0000] [28] [INFO] Waiting for application startup.

This will continue untill I shut it down. This problem weirdly does not happen when I run everything locally.

A difference between local and remote deployment is probably only the port that I set. As you can see, I use port 8081:

[INFO] Listening at: http://0.0.0.0:8081 (1)

I set the port by specifying it in my Kubernetes yaml file, but I am unsure if this has something to do with the problem.

[tiangolo]

WORKERS_PER_CORE is a float telling the container how many workers per core to boot.

If you have it set to 1 and you have 64 processors with 2 cores each one (128 cores), it will start 128 workers.

To make it a fixed number of workers you can use WEB_CONCURRENCY: https://github.com/tiangolo/uvicorn-gunicorn-fastapi-docker#web_concurrency

That will fix the number of workers to that specific number (independent of the number of available cores).

The image is configured to, by default, get the maximum performance available. But it might be a waste of resources and you might not really need to handle 30,000 (or more) requests per second. In that case, you can play around with those configurations as you need.

There are several combinations of configurations you can use, check the docs: https://github.com/tiangolo/uvicorn-gunicorn-fastapi-docker

[tiangolo]

Yeah, I think that could make sense @fdroessler .