06_deploy_api

Lab W05: Deploy API in Google Cloud Platform (GCP)

The goal of this lab is to deploy a basic Python Flask API we created in the previous lab to the Google Cloud Platform (GCP).

1. Prerequisites

1. Have a working version of python
2. Have a working version of Docker Desktop
3. Docker daemon running.

1. Flask API

1.1. Minimal structure of your folder

Inside the folder of this lab, create the following files and folders:

> touch Dockerfile
> mkdir src 
> cd src
> touch requirements.txt
> touch hello.py

You can visualize the structure of your folder with the following command:

> tree

1.2. Content of the files

Inside the requirements.txt file, you should have the following content:

Flask==3.0.3
gunicorn==21.2.0

Inside the hello.py file, you should have the following content:

import os
from flask import Flask

app = Flask(__name__)

@app.route("/")
def hello_world():
    return "Hello World"

if __name__ == "__main__":
    app.run(host="0.0.0.0", port=5001)

Inside Dockerfile add:

FROM python:3.9-slim

# set the environment variable to ensure that the python output is set straight and not buffered
ENV PYTHONUNBUFFERED True

# set the working directory in the container
WORKDIR /src

#copy all files from the current directory to the container
COPY /src/hello.py .
COPY /src/requirements.txt .

ENV PORT 5001

#install the requirements
RUN pip install -r requirements.txt

# Define environment variable to tell the container where the app is (i.e. the file to run)
ENV FLASK_APP=hello.py

# Run hello.py when the container launches
CMD exec gunicorn --bind :$PORT --workers 1 --threads 8 hello:app

In this file, we are using the python:3.9-slim image as the base image. We are also setting the working directory in the container to /src and copying the hello.py and requirements.txt files to the container. We are also installing the requirements and setting the environment variable FLASK_APP to hello.py. Finally, we are running the hello.py file when the container launches. The gunicorn command is used to run the app. e.g. gunicorn --bind :$PORT --workers 1 --threads 8 hello:app means that we are binding the app to the port defined in the environment variable PORT and running the app with 1 worker and 8 threads. hello:app means that we are running the app named app defined in the hello.py file. If you named your Flask app differently in the hello.pyfile, you should replace app with the name you used.

Note : We use gunicorn to run the app. gunicorn is a Python WSGI HTTP Server for UNIX. It is a pre-fork worker model, ported from Ruby's Unicorn project. The Gunicorn server is broadly compatible with various web frameworks, simply implemented, light on server resources, and fairly fast. In our case as we will work with GCP, we will use gunicorn to run the app.

1.3. Test the app locally

Now you can test your app by running the following command:

> docker build -t flask-app .
> docker run -p 5001:5001 flask-app

You should get the following output:

[date-time] [1] [INFO] Starting gunicorn 21.2.0
[date-time] [1] [INFO] Listening at: http://0.0.0.0:5001 (1)
[date-time] [1] [INFO] Using worker: gthread
[date-time] [1] [INFO] Booting worker with pid: 7

And visit the following URL in your web browser: http://localhost:5001/. You should see the message Hello World displayed in your web browser.

omb

2. Create a project in GCP

2.1. Prerequisites

You should, as asked, have created a gmail address for your group and used your coupon to get your 50$ credit on GCP. If you haven't done so, please do it now. Once it is done you should see in the credits section of the billing page of your GCP console that you have 50$ of credit.

1. Go to the GCP console,

2. Create a new project by clicking on the Select a project dropdown,

3. Click on New Project, name your project and click Create,

2.2 Install the Google Cloud SDK

1. Go to the Google Cloud SDK page and follow the instructions to install the SDK for your operating system.

2.2.1 for Mac OS

1. Download the google-cloud-sdk tar file from the Google Cloud SDK page
2. Extract the tar file in your home directory (just double click on the tar file)
3. Open a terminal, you should be in your home directory
4. Run ./google-cloud-sdk/install.sh to install the SDK. You will be prompted to answer a few questions. You can just press Enter to accept the default answers or y to accept the default answers and start the installation.

2.3 Initialize the Google Cloud SDK

1. In your IDE (VScode, Pycharm, etc) or in your terminal, run gcloud init to initialize the SDK. This should appear in your terminal:

Welcome! This command will take you through the configuration of gcloud.

Your current configuration has been set to: [default]

You can skip diagnostics next time by using the following flag:
  gcloud init --skip-diagnostics

Network diagnostic detects and fixes local network connection issues.
Checking network connection...done.                                                                                                                                                                                      
Reachability Check passed.
Network diagnostic passed (1/1 checks passed).

You must log in to continue. Would you like to log in (Y/n)?

2. You will be prompted to log in to your GCP account. The following message should appear in your terminal:

You are logged in as: [google email address].

Pick cloud project to use: 
 [1] your-newly-created-project
 [2] Create a new project
Please enter numeric choice or text value (must exactly match list item):

3. Select the project you just created.

2.4 Change the project

If you want to work in another project, you can see the list of your projects by running the following command in your terminal:

> gcloud projects list

This should give you something like this:

PROJECT_ID                NAME                  PROJECT_NUMBER
something-staging-2587    something-staging     804012817122
something-production-24   something-production  392181605736

Now if you want to change the project you are working on, you can change the project by running the following command in your terminal:

> gcloud config set project something-staging-2587

where something-staging-2587 is the PROJECT_ID of the project you want to work on.

3. Deploy the app to GCP

If you run now the following command in your terminal:

> gcloud run deploy flask-app --region=europe-west1 --source=$(pwd) --allow-unauthenticated

You should get the following output:

Deploying from source requires an Artifact Registry Docker repository to store built containers. A repository named [cloud-run-source-deploy] in region [europe-west1] will be created.

Do you want to continue (Y/n)?  

Accept by typing Y and press Enter.

You should get the following output:

This command is equivalent to running `gcloud builds submit --tag [IMAGE] /Users/matthiaspirlet/info9023-mlops/labs/06_deploy_api` and `gcloud run deploy flask-app --image [IMAGE]`

Building using Dockerfile and deploying container to Cloud Run service [flask-app] in project [project-id] region [europe-west2]
✓ Building and deploying new service... Done.
  ✓ Creating Container Repository...                      
  ✓ Uploading sources...                                
  ✓ Building Container... Logs are available at [https://console.cloud.google.com/cloud-build/builds/...].
  ✓ Creating Revision...                              
  ✓ Routing traffic... 
  ✓ Setting IAM Policy...                               
Done.                                   
Service [flask-app] revision [flask-app-000...] has been deployed and is serving 100 percent of traffic.
Service URL: https://flask-app-...-uc.a.run.app

You can now visit the URL https://flask-app-...-uc.a.run.app in your web browser. You should see the message Hello World displayed in your web browser.

4. Monitor the API logs and usage in Google Cloud

1. In the GCP console, click on the Cloud Run tab. You should see the following screen:
2. Click on the service you just created. You should see the different metrics of your service and above the URL of your service.
3. Click on Logs (or Journaux if you are using the French version of the console) to view the logs

5. Clean up

1. In the GCP console, click on the Cloud Run tab
2. Click on the service you just created
3. Click on Delete to delete the service
4. In the GCP console, click on the Container Registry tab
5. Click on the cloud-run-source-deploy repository
6. Click on the repository you just created and click on Delete to delete the repository
7. In the GCP console, click on the IAM & Admin tab
8. Click on Service accounts and click on the service account you just created. Click on Delete to delete the service account
9. In the GCP console, click on the APIs & Services, then click on Enable APIs and Services
10. Search for Cloud Run API
11. Click on Manage and then click on Disable to disable the API

6. About pricing

Here we built a small app and deployed it to GCP. We used the Cloud Run service which is a serverless platform that allows you to run stateless containers that are invocable via HTTP requests. The pricing of Cloud Run is based on the number of requests and the amount of memory used by the app. You can find more information about the pricing of Cloud Run here. But to give you an idea, in terms of processing power, the first 240k vCPU-seconds per months are free. A vCPU-second is a unit of time that represents the processing power of a single virtual CPU running for one second. After the first 240k vCPU-seconds, you will be charged 0,00001800$ per vCPU-seconde. So for example if you have an app that uses 1 vCPU for 1 second, you will be charged 0,00001800$ after the first 240k vCPU-seconde. This website explains what is a vCPU and how you can calculate vCPU requirements for your app..

In terms of memory, the first 450k Gio-seconde per month are free. After that, you will be charged 0,00000200$ per Gio-seconde. A Gio-seconde is a unit of time that represents the amount of memory used by the app for one second. So for example if you have an app that uses 1 Gio of memory for 1 second, you will be charged 0,00000200$ after the first 450k Gio-seconde.

7. Conclusion

In this lab, you learned how to deploy a basic Hello World! Python Flask API to the Google Cloud Platform (GCP) using the Google Cloud SDK. You also learned how to monitor the API logs and usage in Google Cloud. You also learned how to clean up the resources you created in GCP not to be charged for them 💸💸💸. But you can do way more with more complex Flask APIs and other services of GCP. You can test this tuto and try to deploy more complex APIs that you created in the previous labs.

8. References

• Google Cloud SDK
• Google Cloud Run
• Google Cloud Run API
• Google Cloud Run Pricing
• Google Cloud Run Quickstart
• Deploy a Python Flask Server using Google Cloud Run