Testing GCP Cloud Run triggered by GCP Scheduler

This project is a simple example of how to schedule a Cloud Run's execution. To do so, we dockerize a FastAPI API and deploy it to Cloud Run. Then, we schedule the execution of the Cloud Run's endpoint using Cloud Scheduler.

The sample API here returns a message containing the current time in Unix timestamp format. Two endpoints are available:

• /hoot/unix: returns the current time in Unix timestamp format. Aimed for manual testing;
• /hoot/scheduled: also returns the current time in Unix timestamp format. Aimed for scheduled execution.

Build and Run your image locally

Building your Docker image

To Build your Docker image, run:

docker build -t hoot .

This way you will build a docker image named hoot.

Running your Docker image

To deploy your Docker image run:

docker run -i -p 8080:8080 hoot

This way you will deploy your hoot docker image iteratively on port 8080.

Deploying your project to Google-Cloud-Platform

First, make sure you have access (that is, the APIs are enabled and your permissions/roles are updated) to the following services:

• Artifact Registry
• Cloud Run
• Cloud Scheduler

For this example, also suppose you have:

• GCloud CLI installed and logged in;
• A project available in GCP. You can only deploy stuff in GCP if them are associated with a given project. To this example we will suppose this project is called <YOUR PROJECT NAME>;
• A service account's credentials file authorized. You can authorize your credential file using:

  $ export GOOGLE_APPLICATION_CREDENTIALS=~/path/to/your/file
  $ gcloud auth activate-service-account --key-file $GOOGLE_APPLICATION_CREDENTIALS --project <YOUR PROJECT NAME>

• An Artifact Registry repository. Let's call it <YOUR REPOSITORY NAME>;

Sending your image to Artifact Registry

Now, send your Docker image to the repository you created. To do this, run:

gcloud builds submit --region=us-central1 --tag us-central1-docker.pkg.dev/<YOUR PROJECT NAME>/<YOUR REPOSITORY NAME>/hoot:latest

Deploying your image to Cloud Run

You Docker image is available in the GCP Artifact Registry. Now, you can deploy it to Cloud Run. To do this, run:

gcloud run deploy hoot --image us-central1-docker.pkg.dev/<YOUR PROJECT NAME>/<YOUR REPOSITORY NAME>/hoot:latest --region us-central1 --memory 512Mi --cpu 1 --max-instances 5 --min-instances 0 --port 8080

Turning on the Scheduler

Up to this point, you already have your API running on the endpoint made available by GCP (<YOUR CLOUD RUN URI>). In order to make it be called automatically (let's say, every minute), you need to turn on the Cloud Scheduler. To do this, run:

gcloud scheduler jobs create http hoot-job --schedule "* * * * *" --http-method get --uri <YOUR CLOUD RUN URI SCHEDULER ENDPOINT> --location --us-central1 

Deleting

You test your API and the services. Now you may need to get rid of everything, in order to save unwanted costs. Thus, do:

gcloud scheduler jobs delete hoot-job --location us-central1
gcloud run services delete hoot --region us-central1
gcloud container images delete us-central1-docker.pkg.dev/<YOUR PROJECT NAME>/<YOUR REPOSITORY NAME>/hoot

REFERENCES

• Build and push a Docker image with Cloud Build
• Triggering Cloud Run Jobs with Cloud Scheduler
• Running services on a schedule