Moving from gcloud functions deploy to a production pipeline requires CI/CD, environment separation, and monitoring.

Project Setup 

  gcloud projects create my-project-id
gcloud config set project my-project-id
gcloud services enable cloudfunctions.googleapis.com
gcloud services enable cloudbuild.googleapis.com
gcloud services enable run.googleapis.com

Environment Variables per Stage 

  # Development
gcloud functions deploy my-api-dev \
    --gen2 --runtime=python312 --region=us-central1 \
    --source=. --entry-point=handler \
    --trigger-http --allow-unauthenticated \
    --set-env-vars="ENV=dev,LOG_LEVEL=DEBUG"

# Production
gcloud functions deploy my-api-prod \
    --gen2 --runtime=python312 --region=us-central1 \
    --source=. --entry-point=handler \
    --trigger-http \
    --set-env-vars="ENV=prod,LOG_LEVEL=INFO" \
    --set-secrets="API_KEY=prod-api-key:latest" \
    --min-instances=1 \
    --max-instances=50

CI/CD with Cloud Build 

  # cloudbuild.yaml
steps:
  - name: 'python:3.12'
    entrypoint: pip
    args: ['install', '-r', 'requirements.txt', '-t', './package']

  - name: 'python:3.12'
    entrypoint: python
    args: ['-m', 'pytest', 'tests/', '-v']

  - name: 'gcr.io/google.com/cloudsdktool/cloud-sdk'
    entrypoint: gcloud
    args:
      - functions
      - deploy
      - my-api
      - --gen2
      - --runtime=python312
      - --region=us-central1
      - --source=.
      - --entry-point=handler
      - --trigger-http

Trigger on git push:

  gcloud builds triggers create github \
    --repo-name=my-repo --repo-owner=my-org \
    --branch-pattern="^main$" \
    --build-config=cloudbuild.yaml

Custom Domain & HTTPS

Map a custom domain via Cloud Run (Gen 2 functions run on Cloud Run):

  gcloud run domain-mappings create \
    --service=my-api \
    --domain=api.example.com \
    --region=us-central1

Add the DNS records shown in the output to your domain registrar.

Monitoring Dashboard

Key Cloud Monitoring metrics:

  • Request count and latency (p50, p95, p99)
  • Error rate (4xx, 5xx)
  • Active instances and cold start frequency
  • Memory and CPU utilization

Create an alert policy for error rate > 5% over 5 minutes.

Rollback 

  # List revisions
gcloud functions describe my-api --gen2 --region=us-central1

# Route 100% traffic to previous revision
gcloud run services update-traffic my-api \
    --to-revisions=PREVIOUS_REVISION=100 \
    --region=us-central1

Cost Management

Strategy Impact
Set --max-instances Cap runaway scaling costs
Use --min-instances=0 in dev No idle cost
Right-size memory Faster = potentially cheaper
Use Cloud Scheduler + batch Reduce invocation count

Production Cloud Functions on GCP leverage Cloud Run infrastructure for reliability and scale.