🚀

Cloud Resources Setup

⏱️ 15 minutes

Phase 1 of 7

ℹ️ Objective

Set up cloud infrastructure and development environment for MLOps pipeline

☁️

Google Cloud Platform

Recommended for Kubeflow

✅ Best Kubeflow support

✅ Managed Kubernetes (GKE)

✅ GPU instances available

✅ $300 free credits

Start GCP Setup →

☁️ Google Cloud Platform Setup

⏱️

Estimated Time: 5-10 minutes

GCP offers the best Kubeflow support and managed Kubernetes. Perfect for production MLOps workflows.

1 Prerequisites & Account Setup

What you need:

✓ Google Cloud account - Free tier available with $300 credits
✓ gcloud CLI - Command-line interface for Google Cloud
✓ kubectl - Kubernetes command-line tool
✓ Credit card - For verification only (won't be charged)

Quick setup:

Install gcloud CLI

    
      
# Install gcloud CLI (macOS)
curl https://sdk.cloud.google.com | bash

# Install gcloud CLI (Linux)
curl https://sdk.cloud.google.com | bash

# Install gcloud CLI (Windows)
# Download from: https://cloud.google.com/sdk/docs/install

2 Create GCP Project

Create a new project to organize your MLOps resources and enable billing.

Create and Configure Project

    
      
# Create new project
gcloud projects create mlops-workshop --name="MLOps Workshop"

# Set project as default
gcloud config set project mlops-workshop

# Enable required APIs
gcloud services enable container.googleapis.com
gcloud services enable compute.googleapis.com
gcloud services enable storage.googleapis.com
gcloud services enable aiplatform.googleapis.com
gcloud services enable cloudbilling.googleapis.com
gcloud services enable gkehub.googleapis.com

# Verify APIs are enabled
gcloud services list --enabled
gcloud container fleet create --display-name="mlops-fleet"

ℹ️ 💡 Pro Tip

Replace the project name with something unique to you. The $(date +%s) adds a timestamp to make it unique.

3 Create Google Kubernetes Engine (GKE) Cluster

Create a managed Kubernetes cluster that will host your MLOps workloads.

Create GKE Cluster

    
      
# Create GKE cluster
gcloud container clusters create mlops-cluster \
  --zone=us-central1-a \
  --num-nodes=3 \
  --machine-type=e2-standard-4 \
  --enable-autoscaling \
  --min-nodes=3 \
  --max-nodes=5 \
  --enable-autorepair \
  --enable-autoupgrade \
  --disk-size=50 \
  --disk-type=pd-standard \
  --enable-fleet

# This will take 5-10 minutes to complete

Cluster Configuration:

• 3 nodes with 4 vCPUs and 16GB RAM each (e2-standard-4)
• Auto-scaling enabled (1-5 nodes)
• Auto-repair and auto-upgrade enabled
• 50GB standard persistent disk per node

4 Connect to Your Cluster

Get cluster credentials and verify your connection.

Connect to Cluster

    
      
# Install GKE Auth Plugin
gcloud components install gke-gcloud-auth-plugin -q

# Get cluster credentials
gcloud container clusters get-credentials mlops-cluster --zone=us-central1-a

# Verify kubectl
kubectl version --client

# Verify connection
kubectl get nodes

# Check cluster info
kubectl cluster-info

Expected Output:

You should see 3 nodes listed with status "Ready". If you see any issues, wait a few minutes for the cluster to fully initialize.

5 Verify Setup

Run these commands to ensure everything is working correctly.

Verification Commands

    
      
# Check cluster status
kubectl get nodes -o wide

# Check available resources
kubectl top nodes

# Test kubectl connectivity
kubectl run test-pod --image=nginx --restart=Never
kubectl get pods
kubectl delete pod test-pod

✅ 🎉 GCP Setup Complete!

Your GCP environment is ready! You can now proceed to install Kubeflow in the next phase.

6 Cleanup (Optional)

When you're done with the workshop, you can delete the cluster to avoid charges.

Delete GKE Cluster

    
      
# Delete the cluster (this will remove all data)
gcloud container clusters delete mlops-cluster --zone=us-central1-a

# Confirm deletion when prompted
# This will take 5-10 minutes to complete

⚠️ Important:

• This will permanently delete your cluster and all data
• You'll need to recreate everything if you want to continue
• Only run this when you're completely done with the workshop

← Back to Dashboard Next: Pipelines →