dradamroe / cloud-variations Goto Github PK

There are some typos in the tutorial at:
https://github.com/DrAdamRoe/cloud-variations#install-locally

• Install packages: pip install -r requirements.tst

should be:

• Install packages: pip install -r requirements.txt

• run your flask application:

flask run

Should be (at least in my case):

• run your flask application:

FLASK run

Many students are using FastAPI and it seems mature enough. In addition to be a more contemporary framework (e.g. async), the code is a little simpler than Flask and slightly more readable than something similar from Flask. This switch to be a good idea for this tutorial.

note: be careful of Python version issues, as usual. The official docs note that Python 3.9 is the latest supported version, although Google Functions recommends using 3.11 as the application runtime. For this tutorial, it is important that the Python version is stable: better to update the SDK and use Python 3.9 throughout than use Python 3.11 as the application runtime but 3.9 for the Google Cloud CLI.

The deployments here succeeded but the runtime was unable to start properly. The result is that the server threw an error (a 500 in this case) and the pipeline showed as successful, since it received a valid HTTP response from the server, it just wasn't a 200 or the message we expected. To fix, the response of the server should be checked.

Python 3.11 is currently recommended for Google Functions

Information here and here

There might be a viable alternative now to the full nginx setup by hand.

…especially if it works on windows.

Currently, switching to a new project (e.g. at the start of a new semester) requires manually setting up a new project, which includes getting all of the correct gcloud apis enabled and correct permissions set up on service accounts. This is somewhat documented in the github workflows yaml file, because the issues generally appear in the ci/cd process and not in the manual instructions at the command line. The best way to automate this is probably to use terraform, e.g. Google Project Service.