File: mlops-service/lambda_function.py

"""
AWS Lambda handler for Flask MLOps Service
Converts Flask WSGI app to Lambda-compatible function
"""
import serverless_wsgi
from app import app

def lambda_handler(event, context):
    """
    AWS Lambda handler function

    Args:
        event: API Gateway request event
        context: Lambda execution context

    Returns:
        API Gateway response format
    """
    return serverless_wsgi.handle_request(app, event, context)

Add this line to your existing requirements.txt:

# Existing dependencies
flask==3.0.0
flask-cors==4.0.0
prometheus-client==0.19.0
psycopg2-binary==2.9.9
python-dotenv==1.0.0

# Add for Lambda deployment
serverless-wsgi==3.0.3

Lambda needs a ZIP file containing:

• Your application code (app.py, lambda_function.py)
• All Python dependencies (Flask, prometheus-client, etc.)
• Must be structured correctly for Lambda to find the handler

Mac/Linux:

# Navigate to mlops-service directory
cd mlops-service

# Create a clean directory for the package
mkdir -p lambda-package
cd lambda-package

# Install dependencies into this directory
pip install -r ../requirements.txt -t .

# Copy application files
cp ../app.py .
cp ../lambda_function.py .

# Create ZIP file
zip -r ../lambda-deployment.zip .

# Go back to mlops-service directory
cd ..

# Verify ZIP was created
ls -lh lambda-deployment.zip

Windows (PowerShell):

# Navigate to mlops-service directory
cd mlops-service

# Create a clean directory for the package
New-Item -ItemType Directory -Force -Path lambda-package
cd lambda-package

# Install dependencies into this directory
pip install -r ..\requirements.txt -t .

# Copy application files
Copy-Item ..\app.py .
Copy-Item ..\lambda_function.py .

# Create ZIP file (requires PowerShell 5.0+)
Compress-Archive -Path .\* -DestinationPath ..\lambda-deployment.zip -Force

# Go back to mlops-service directory
cd ..

# Verify ZIP was created
Get-Item lambda-deployment.zip

Access AWS Lambda:

1. Sign in to AWS Console
2. Search for "Lambda" in the search bar
3. Click "Lambda" to open the Lambda console
4. Make sure you're in the same region as Lab 7 (e.g., us-east-1)

Create a new function:

1. Click "Create function" button
2. Select "Author from scratch"
3. Configure:
   • Function name: mlops-service-lambda
   • Runtime: Python 3.11
   • Architecture: x86_64
   • Permissions: Create a new role with basic Lambda permissions
4. Click "Create function"

Upload your ZIP file:

1. In the function page, scroll to "Code source" section
2. Click "Upload from" dropdown
3. Select ".zip file"
4. Click "Upload"
5. Select your lambda-deployment.zip file
6. Click "Save"

Set the Lambda handler:

1. Scroll to "Runtime settings" section
2. Click "Edit"
3. Set Handler to: lambda_function.lambda_handler
4. Click "Save"

Add your environment variables:

1. Click "Configuration" tab
2. Click "Environment variables" in left sidebar
3. Click "Edit"
4. Add variables (click "Add environment variable" for each):

Adjust function settings:

1. Stay in "Configuration" tab
2. Click "General configuration" in left sidebar
3. Click "Edit"
4. Set:
   • Memory: 512 MB (enough for Flask + Prometheus)
   • Timeout: 30 seconds (longer than default 3 seconds)
5. Click "Save"

Flow:

User Request (https://your-api.execute-api.us-east-1.amazonaws.com/prod/health)
    ↓
API Gateway (receives HTTP request)
    ↓
Lambda Function (processes request)
    ↓
API Gateway (returns HTTP response)
    ↓
User (receives response)

From Lambda console:

1. In your Lambda function page, click "Add trigger"
2. Select "API Gateway"
3. Configure:
   • API type: HTTP API (simpler than REST API)
   • Security: Open (we'll add security later if needed)
4. Click "Add"

Copy your API endpoint:

1. In the "Triggers" section, click on the API Gateway trigger
2. You'll see "API endpoint" URL like:

Test the health endpoint:

# Replace with your actual API Gateway URL
curl https://YOUR_API_GATEWAY_URL

# Expected: Health check JSON response

Test the /health endpoint specifically:

curl https://YOUR_API_GATEWAY_URL/health

Expected Response:

{
  "status": "healthy",
  "service": "mlops-service-prometheus",
  "timestamp": "2024-01-15T10:30:00.000000",
  "monitoring": "prometheus",
  "metrics_endpoint": "/metrics",
  "environment": "production"
}

Test Prometheus metrics:

curl https://YOUR_API_GATEWAY_URL/metrics

You should see Prometheus metrics output:

# HELP python_info Python platform information
# TYPE python_info gauge
python_info{implementation="CPython",major="3",minor="11"...} 1.0
...

Switch from EC2 to Lambda:

1. Go to vercel.com
2. Click on your project
3. Go to Settings → Environment Variables
4. Find MLOPS_SERVICE_URL
5. Click "Edit"
6. Update to your API Gateway URL:

Redeploy to use new environment variable:

1. Go to Deployments tab
2. Click on the latest deployment
3. Click "Redeploy" button
4. Wait for deployment to complete

Test the complete serverless flow:

1. Visit your Vercel URL (e.g., https://your-app.vercel.app)
2. Create a new business or use existing
3. Open the chat interface
4. Send a message to the AI
5. Check if metrics are tracked

Verify metrics on Lambda:

# Check metrics endpoint
curl https://YOUR_API_GATEWAY_URL/metrics | grep ai_requests_total

# Should show incremented counter

Check Lambda CloudWatch logs:

1. Go to Lambda console
2. Click on your function
3. Click "Monitor" tab
4. Click "View CloudWatch logs"
5. Click on the latest log stream

You should see:

• Incoming requests
• Prometheus metrics updates
• Any errors or warnings

Test EC2 response time:

# Time 10 requests to EC2
for i in {1..10}; do
  time curl -s http://YOUR_EC2_IP:5001/health > /dev/null
done

Test Lambda response time:

# Time 10 requests to Lambda
for i in {1..10}; do
  time curl -s https://YOUR_API_GATEWAY_URL/health > /dev/null
done

Warm Starts:

After the first request, Lambda keeps the environment "warm" for ~15 minutes. Subsequent requests are fast (~50-150ms).

Decision Time:

To restart later:

# Instance state → Start instance
# Get new public IP (changes after stop/start)
# Update Vercel MLOPS_SERVICE_URL if switching back

Check Lambda usage:

1. Lambda console → Functions
2. Click "Monitor" tab
3. View invocations, duration, errors

Check EC2 usage:

1. EC2 console → Instances
2. Check instance hours used

Check free tier usage:

1. AWS Console → Billing
2. Free Tier → View usage
3. See Lambda requests and EC2 hours remaining

Create EventBridge Rule:

1. Go to AWS Console → EventBridge
2. Create rule → Schedule
3. Configure:
   • Name: mlops-lambda-warmer
   • Schedule: Rate expression: rate(5 minutes)
4. Target: Lambda function → mlops-service-lambda
5. Create